CN107808333A - A kind of commodity launch decision system, method and device - Google Patents

Abstract

This application discloses a kind of commodity to launch decision system, method, apparatus and electronic equipment.Wherein, the commodity launch decision system and include launching decision-making subsystem and launch staining effect subsystem；Compared with the actual dispensing effect data got can be launched effect data by the staining effect subsystem with prediction, the gap of two kinds of data is obtained, and the gap is fed back into dispensing decision-making subsystem；The decision-making subsystem of launching can carry out adaptive adjustment according to the gap, and treat that decision-making items list carries out dispensing decision-making using the dispensing decision-making subsystem after adjustment to be next.The commodity provided using the application launch decision system, can launch the gap of effect data with prediction according to actual dispensing effect data, adjust launch decision-making subsystem in real time, so as to reach the effect for improving commodity and launching decision accuracy.

Description

A kind of commodity launch decision system, method and device

Technical field

The application is related to machine learning techniques field, and in particular to a kind of commodity launch decision system；Corresponding to above-mentioned system System, the application are related to a kind of commodity and launch decision-making technique, device and electronic equipment simultaneously.

Background technology

With the continuous development and popularization of machine learning techniques, increasing commodity launch decision business and use digitization Operation way substitutes the artificial method of operation and instructed launching decision process, i.e.,：Manually done by launching decision system replacement Decision-making is to improve service operation efficiency.Below to above-mentioned digitization operation way and the party by taking poly- commodity auction business to one's profit as an example The problem of formula is present is briefly described.

Commodity under poly- commodity auction business to one's profit are launched decision process and comprised the following steps：1) target volume in value etc. is referred to Mark carries out overall planning；2) according to integrated planning to may participate in the affiliated commodity classification of lipuid goods, each commodity classification has The parameter such as resource bit quantity carry out physical planning；3) businessman is the commodity registration of auction to be participated in；4) registration commodity are carried out Examination ＆ verification, it is determined to the commodity of activity.In traditional artificial operation way, above-mentioned steps be required to by operation personnel with Manual type is handled.After digitization operation way, above-mentioned 4th step is by the dispensing decision-making based on machine learning techniques System is handled, i.e.,：Decision system is launched by commodity to assess the commodity of registration automatically, to determine finally participate in The poly- commodity for calculating activity.After being runed using digitization, the part that operation personnel is instead of by dispensing decision system works, thus Improve service operation efficiency.

After the dispensing result of decision progress commodity dispensing that decision system provides is launched, during service operation, need Periodically the actual dispensing effect of commodity assessed, and be adjusted according to assessment result to launching decision system, to protect The realization of card operation target.For example, when if actual sales revenue caused by poly- activity to one's profit is not reaching to target volume in value, need Adjust the relevant parameter for launching decision system.

At present, the adjustment work of above-mentioned dispensing recruitment evaluation, dispensing decision system parameter is responsible for place by operation personnel Reason, it is seen then that launch the level that the guidance that decision system is launched to commodity only rests on data target prediction, do not reach complete The effect of digitization operation (alternatively referred to as " wisdom operation "), thus, launch the auxiliary operation meaning that decision system is launched to commodity Justice is extremely limited.In addition, it is completely mutually solely under the prior art, between the different decision systems under different decision-making scenes Vertical, unrelated decision system construction cost height and poor expandability.

In summary, commodity under prior art launch decision system have that decision accuracy is low and extension cost is high asks Topic.

The content of the invention

The application provides a kind of commodity and launches decision system, and decision system presence is launched to solve the commodity under prior art Decision accuracy is low and extends the problem of cost is high.The application provides a kind of commodity dispensing decision-making technique, device and electronics and set in addition It is standby.

The application provides a kind of commodity and launches decision system, including：Launch decision-making subsystem and launch staining effect subsystem System；

The dispensing decision-making subsystem, for receiving the decision requests of the items list for treating decision-making, and according to each The characteristic of commodity and the dispensing decision model previously generated, it is determined that the commodity that can be launched；And receive the dispensing effect The actual dispensing effect data for the commodity launched that feedback subsystem provides is relative to the pre- of the commodity launched The first gap for launching effect data is surveyed, and is adaptively adjusted according to first gap, and uses the dispensing after adjustment Decision-making subsystem treats that decision-making items list carries out dispensing decision-making to next；

The dispensing staining effect subsystem, for obtaining the actual dispensing effect data of the commodity that can be launched, and The prediction of the actual dispensing effect data of the commodity launched and the commodity launched is launched into effect data to carry out Compare, obtain first gap；And first gap is sent out.

Optionally, it is described adaptively to be adjusted according to first gap, including：

Adjust the decision rule of the dispensing decision-making subsystem；

Adjust the boundary condition of the dispensing decision-making subsystem；

Adjust the constraints of the dispensing decision-making subsystem；

Adjust the processing parameter of the dispensing decision-making subsystem；

The adjustment for launching decision model is carried out according to the actual effect data of launching.

Optionally, in addition to：Category planning forecast subsystem；

The category plans subsystem, for receiving for the specific category planning request for launching effect target, and according to The characteristic of each commodity classification to be planned, the specific category plan model launched effect target and previously generated, Generate category planning forecast result；And the category planning forecast result is sent out；And receive described launch and imitate The category that fruit feedback subsystem provides plans second gap of the actual result relative to category planning forecast result, and according to described Second gap is adaptively adjusted, and is carried out category next time using the category planning subsystem after adjustment and planned；Wherein, institute State category planning actual result to refer to, caused reality after the dispensing decision-making of commodity is carried out according to the category planning forecast result Category program results；

The dispensing staining effect subsystem, it is additionally operable to obtain the category planning actual result, and the category is advised Actual result is drawn compared with the category planning forecast result, obtains second gap；And by second gap It is sent out；

The dispensing decision-making subsystem, it is additionally operable to receive the category planning forecast that the category planning subsystem provides As a result, the commodity that and according to determining the category planning forecast result can be launched.

Optionally, in addition to：Invite decision-making subsystem；

The invitation decision-making subsystem, the category planning forecast knot provided for receiving the category planning subsystem Fruit, and according to the characteristics of the commodity for meeting the category planning forecast result and the invitation decision model previously generated, really Surely the commodity that can be invited；And receive the actual throwing for launching the commodity invited that staining effect subsystem provides The 3rd gap that effect data launches effect data relative to the prediction of the commodity invited is put, and it is poor according to the described 3rd Away from adaptively being adjusted, and invitation decision-making is carried out to commodity using the invitation decision-making subsystem after adjustment；

The dispensing staining effect subsystem, the actual dispensing effect data of commodity that can be invited described in obtaining is additionally operable to, And effect data is launched into the prediction of the actual dispensing effect data of the commodity invited and the commodity invited and entered Row compares, and obtains the 3rd gap；And the 3rd gap is sent out.

Optionally, in addition to：Launch effect target prediction subsystem；

The dispensing effect target prediction subsystem, for receiving the dispensing effect target prediction for the specific dispensing cycle Request, and thrown according to the specific characteristic for launching the cycle and the dispensing effect target prediction model previously generated, generation Put effect prediction target；And the dispensing effect prediction target is sent out；And receive the dispensing staining effect The dispensings effect realistic objective that subsystem provides is relative to the 4th gap for launching effect prediction target, and according to described the Four gaps are adaptively adjusted, and carry out target prediction next time using the dispensing effect target prediction subsystem after adjustment；

The dispensing staining effect subsystem, it is additionally operable to obtain the dispensing effect realistic objective, and described launch is imitated Fruit realistic objective obtains the 4th gap compared with the dispensing effect prediction target；And by the 4th gap It is sent out.

Optionally, in addition to：Model evaluation subsystem；

The model evaluation subsystem, commodity are launched for receiving the assessment request for particular model, and according to actual And corresponding actual dispensing effect data, generate forecast assessment result；And by the forecast assessment result to outgoing Send；And the actual assessment result that the offer of staining effect subsystem is provided is received relative to the forecast assessment result 5th gap, and adaptively adjusted according to the 5th gap, and using the model evaluation subsystem after adjustment to next Individual particular model is assessed；

The dispensing staining effect subsystem, is additionally operable to obtain the actual assessment result, and by the actual assessment knot Fruit obtains the 5th gap compared with the forecast assessment result；And the 5th gap is sent out.

Optionally, in addition to：Data mining subsystem；

The data mining subsystem includes：The first data mining subsystem for only being had an impact to specific decision-making scene and The second data mining subsystem that multiple decision-making scenes are had an impact；

First data mining subsystem, for the number of extraction first from the commodity related data of specific decision-making scene According to；

Second data mining subsystem, for extracting the second data from the commodity related data of enterprise-level.

Optionally, in addition to：Run Monitor And Control Subsystem；

The operation Monitor And Control Subsystem, the decision process of decision system is launched for monitoring the commodity.

Optionally, in addition to：Extend subsystem；

The extension subsystem, determined for newly-increased decision-making subsystem or predicting subsystem to be increased into the commodity dispensing Plan system.

Accordingly, a kind of commodity itself are please also provided and launch decision-making technique, including：

Obtain the items list for treating decision-making；

Decision system is launched by current commodity each commodity for treating decision-making are carried out with dispensing decision-making, obtained and launch decision-making knot Fruit；

Obtain the actual dispensing effect data after commodity are launched according to the dispensing result of decision；

The actual effect data of launching is launched with the generation current commodity for launching result of decision institute foundation The prediction that decision system provides is launched effect data and is compared, and obtains the actual effect data of launching relative to the prediction Launch the gap of effect data；

Commodity are carried out according to the gap and launch decision system adjustment；

After the completion of the commodity launch decision system adjustment, the commodity of the renewal of acquisition are launched into decision system as current Commodity launch decision system.

Optionally, the dispensing effect includes many-sided dispensing effect；

It is described that the reality is launched into effect data compared with the prediction dispensing effect data, using such as lower section Formula：

Each side is launched into the actual dispensing effect data of effect and effect number is launched in the corresponding prediction According to being compared respectively；

Accordingly, the gap includes many-sided gap for launching effect；

It is described that commodity dispensing decision system adjustment is carried out according to the gap, in the following way：

The gap that effect is launched according to each side carries out commodity dispensing decision system adjustment.

Optionally, it is described that commodity dispensing decision system adjustment is carried out according to the gap, including following method of adjustment is extremely Few one：

Adjust the decision rule that commodity launch decision system；

Adjust the boundary condition that commodity launch decision system；

Adjust the constraints that commodity launch decision system；

Adjust the processing parameter that commodity launch decision system；

According to newly-increased actual dispensing effect data launch the adjustment of decision model.

Optionally, the actual dispensing effect data obtained after commodity are launched according to the dispensing result of decision, is used Following manner：

By data base querying, the actual dispensing effect data is obtained.

Optionally, the current commodity launches decision system and carries out the characteristic of decision-making institute foundation including only being determined to specific Fisrt feature data that plan scene has an impact and the second feature data that multiple decision-making scenes are had an impact.

Optionally, the characteristic that the current commodity dispensing decision system carries out decision-making institute foundation includes the spy of Knowledge Levy data.

Optionally, it is described that dispensing decision-making is carried out to each commodity for treating decision-making by current commodity dispensing decision system, adopt Use following manner：

According to the category planning forecast result previously generated, dispensing decision-making is carried out to each commodity for treating decision-making.

Optionally, the category planning forecast result, is generated using following steps：

Receive the category planning request for specific dispensing effect target；

Obtain the characteristic of each commodity classification to be planned；

According to the specific dispensing effect target, the characteristic of the commodity classification and the category planning mould previously generated Type, generate the category planning forecast result.

Optionally, the specific dispensing effect target, is generated using following steps：

Receive the dispensing effect target prediction request for the specific dispensing cycle；

Obtain the specific characteristic for launching the cycle；

Thrown according to the specific characteristic for launching the cycle and the dispensing effect target prediction model previously generated, generation The predicted value of effect target is put, as the specific dispensing effect target.

Optionally, in addition to：

According to the gap, adjust the current commodity and launch the commodity that decision system launch decision-making institute foundation Feature and the current commodity launch decision system use decision making algorithm at least one.

Accordingly, a kind of commodity itself are please also provided and launch decision making device, including：

Items list acquiring unit, the items list of decision-making is treated for obtaining；

Decision package is launched, each commodity for treating decision-making are launched certainly for launching decision system by current commodity Plan, obtain and launch the result of decision；

It is actual to launch effect data acquiring unit, for obtaining the reality after commodity are launched according to the dispensing result of decision Launch effect data；

Contrast on effect unit, for the reality to be launched into effect data with generating dispensing result of decision institute foundation The current commodity is launched the prediction dispensing effect data that decision system provides and is compared, and obtains the reality and launches effect number According to the gap that effect data is launched relative to the prediction；

System call interception unit, decision system adjustment is launched for carrying out commodity according to the gap；

System update unit, after the completion of launching decision system adjustment for the commodity, the commodity of the renewal of acquisition are thrown Put decision system and launch decision system as current commodity.

Accordingly, itself a kind of electronic equipment is please also provided, including：

Processor；And

Memory, for storing the program for realizing that commodity launch decision-making technique, the equipment is powered and runs the commodity and launches After the program of decision-making technique, following step is performed：Obtain the items list for treating decision-making；Decision system pair is launched by current commodity Each commodity for treating decision-making carry out dispensing decision-making, obtain and launch the result of decision；Obtain and business is launched according to the dispensing result of decision Actual dispensing effect data after product；By the actual effect data of launching with generating the institute for launching result of decision institute foundation State the prediction dispensing effect data that current commodity dispensing decision system provides to be compared, obtain the actual dispensing effect data The gap of effect data is launched relative to the prediction；Commodity are carried out according to the gap and launch decision system adjustment；The business After the completion of product launch decision system adjustment, the commodity of the renewal of acquisition are launched into decision system and launch decision-making system as current commodity System.

Compared with prior art, the commodity that the application provides, which launch decision system, to be included launching decision-making subsystem and launching imitating Fruit feedback subsystem；Wherein, the staining effect subsystem can launch the actual dispensing effect data got and prediction Effect data is compared, and obtains the gap of two kinds of data, and the gap is fed back into dispensing decision-making subsystem；The dispensing is determined Plan subsystem can carry out adaptive adjustment according to the gap, and be treated using the dispensing decision-making subsystem after adjustment to be next Decision-making items list carries out dispensing decision-making.

The commodity provided using the application launch decision system, can launch effect with prediction according to actual effect data of launching Decision-making subsystem is launched in the gap of fruit data, in real time adjustment；This processing mode, commodity can be effectively improved and launch determining for decision-making Plan precision.

In the commodity that the application provides launch decision system, the subsystem of different decision-making scenes can be shared by the second data Second data of subsystem offer are provided, share staining effect subsystem, merging run Monitor And Control Subsystem etc.；Therefore, the application The commodity of offer launch decision system, can effectively improve the scalability of system.

In addition, in the commodity that the application provides launch decision system, may also include launch effect target prediction subsystem, Category plans subsystem and model evaluation subsystem, and different decision-making subsystems provide the result of decision of different aspect, by multiple Sub- decision-making link ultimately forms the dispensing result of decision；This processing mode so that multiple coherent under whole dispensing decision business Scene decision-making subsystem and staining effect subsystem between be configured to a big closed loop；Therefore, the commodity that the application provides Decision system is launched, the operation way of complete data can be realized, i.e.,：It can reach the operational effect of wisdom.

Brief description of the drawings

Fig. 1 is the schematic diagram for the embodiment that a kind of commodity that the application provides launch decision system；

Fig. 2 is the first specific schematic diagram of the embodiment that a kind of commodity that the application provides launch decision system；

Fig. 3 is the second specific schematic diagram of the embodiment that a kind of commodity that the application provides launch decision system；

Fig. 4 is that the specific of the data mining subsystem for the embodiment that a kind of commodity that the application provides launch decision system is shown It is intended to；

Fig. 5 is the 3rd specific schematic diagram of the embodiment that a kind of commodity that the application provides launch decision system；

Fig. 6 is the 4th specific schematic diagram of the embodiment that a kind of commodity that the application provides launch decision system；

Fig. 7 is the 5th specific schematic diagram of the embodiment that a kind of commodity that the application provides launch decision system；

Fig. 8 is the 6th specific schematic diagram of the embodiment that a kind of commodity that the application provides launch decision system；

Fig. 9 is the flow chart for the embodiment that a kind of commodity that the application provides launch decision-making technique；

Figure 10 is the schematic diagram for the embodiment that a kind of commodity that the application provides launch decision making device；

Figure 11 is the schematic diagram of the embodiment for a kind of electronic equipment that the application provides.

Embodiment

Many details are elaborated in the following description in order to fully understand the application.But the application can be with Much it is different from other manner described here to implement, those skilled in the art can be in the situation without prejudice to the application intension Under do similar popularization, therefore the application is not limited by following public specific implementation.

In this application, there is provided a kind of commodity launch decision system, and a kind of commodity launch decision-making technique, device and Electronic equipment.It is described in detail one by one in the following embodiments.

Fig. 1 is refer to, it launches the schematic diagram of the embodiment of decision system for the commodity of the application.The commodity are launched and determined Plan system, including：Launch decision-making subsystem 101 and launch staining effect subsystem 103.

The dispensing decision-making subsystem 101, for receiving the decision requests of the items list for treating decision-making, and according to each The characteristic of individual commodity and the dispensing decision model previously generated, it is determined that the commodity that can be launched；And receive described launch and imitate The actual dispensing effect data for the commodity launched that fruit feedback subsystem 103 provides is relative to the commodity launched Prediction launch the first gap of effect data, and adaptively adjusted according to first gap, and using adjusting after Launch decision-making subsystem and treat that decision-making items list carries out dispensing decision-making to next.

The items list for treating decision-making includes at least one commodity for treating decision-making.The commodity for treating decision-making have multiple The characteristic of dimension, the characteristic of the commodity include at least one of following any characteristic：History conclusion of the business effect Characteristic, the characteristic of history buyer's behavior, the characteristic of commercial quality.

The characteristic of the history conclusion of the business effect, mainly summed up from the daily record of record commodity transaction information anti- The characteristic of the history conclusion of the business effect of commodity is reflected, for example, the total business volume of commodity, commodity launch the cycle in each history Consumption sum etc..

The characteristic of buyer's behavior, the reflection business mainly summed up from the daily record of record buyer's behavioral data The characteristic of buyer's behavior of product, for example, browsing the number of users of the commodity in one month in the past, buying the number of users of the commodity Deng.

Buyer's behavior includes at least one of following any buyer's behavior：Navigation patterns, click on behavior, collection row To add shopping cart behavior.

The characteristic of the commercial quality can generate in several ways, for example, can be the commodity matter manually set Data are measured, or commercial quality data obtained etc. are calculated by commercial quality forecast model.The commercial quality predicts mould Type refers to, by machine learning algorithm, the model obtained from training sample focusing study.

For example, under poly- commodity auction business to one's profit, the characteristic of commodity may include：Past 3 days, one week, one month The number of users that inside browses the commodity, the number of users for buying the commodity, the commodity are added into the number of users of shopping cart, collect the commodity Number of users, the total business volume of the commodity, the commodity each history launch the cycle consumption sum；In addition, the commodity Characteristic may also include：The commodity are in the conclusion of the business accounting of whole market, and the various data in one month same period refer in the past The data such as mark.

The characteristic of the commodity, it can be extracted and obtained from commodity related data by data mining subsystem, for example, from The characteristic that the history conclusion of the business effect is excavated in the daily record of commodity transaction information is recorded, from record buyer's behavioral data Characteristic of buyer's behavior etc. is excavated in daily record.

The dispensing decision-making subsystem 101 is being received for after treating the decision requests of items list of decision-making, by root According to the characteristic of each commodity, it is predicted by dispensing effect of the dispensing decision model of the subsystem to each commodity, And effect data is launched according to prediction and determines which commodity can be launched, i.e.,：Formed and launch the result of decision；Finally, will can launch The result of decision, which is shown in, to be regularly updated in the merchandise display page for launching commodity.

The dispensing result of decision is (i.e.：Determine launch commodity) be the items list for treating decision-making subset.Institute State dispensing decision model to refer to, by machine learning algorithm, from the various features data of commodity and the true dispensing effect of commodity The focusing study of history corresponding record obtain dispensing decision model.The history corresponding record collection includes that there is actual launch to imitate The positive sample of fruit data and the negative sample without actual dispensing effect data.

After the dispensing result of decision is formed, will launch the result of decision includes the dispensing decision-making subsystem 101 that the application provides Commodity be shown in regularly update launch commodity the merchandise display page in.Different commodity are corresponding different in the merchandise display page Resource-niche (be commonly called as：Cheat position), and commodity occupancy resource-niche has the term of validity；Launched when launching decision-making subsystem 101 to be follow-up After the new dispensing result of decision of period-producer, by according to each resource-niche in the new dispensing result of decision renewal merchandise display page The commodity of displaying.User can click on its commodity resource-niche interested, and then browse the money when checking the merchandise display page Source position shows the activity details of commodity, so that user buys the commodity.

The decision-making subsystem 101 of launching realizes that the items list for treating decision-making launch the function of decision-making.To be clear Intuitively illustrate that the commodity that the application provides launch decision system, launch decision-making below to support to gather the commodity for calculating auction business Decision system is launched exemplified by system to the commodity that the application provides to illustrate.

The commodity of poly- auction business to one's profit are supported to launch decision system and include：Launching decision-making subsystem, (also referred to as commodity are examined Core decision-making subsystem), data mining subsystem and launch staining effect subsystem.Wherein, decision-making subsystem reception is launched to be directed to Participate in the dispensing decision requests of multiple commodity of poly- auction business to one's profit；Data mining subsystem is from commodity related data The characteristic of each registration commodity is excavated, for example, commercial quality, businessman's prestige, commodity are added into the quantity of shopping cart, business Product are added into the core feature data such as the number that the quantity of collection, the sales volume of commodity, commodity are browsed, commodity evaluation, And the non-core characteristic such as commodity price, merchandise discount；Launch decision-making subsystem and data mining subsystem offer is provided Each registration commodity characteristic after, according to it is each registration commodity characteristic, pass through commodity launch decision-making subsystem System commodity examination ＆ verification model to it is each registration commodity dispensing effect (such as：Resource-niche sales volume etc.) it is predicted, and according to pre- Dispensing effect data is surveyed to determine that the poly- commodity for calculating activity can be participated in.

The commodity that the application provides launch decision system, emphasize it is according to the feedback of dispensing staining effect subsystem 103, It is described it is actual launch effect data and prediction launch gap between effect data, the dispensing decision-making subsystem is carried out it is adaptive It should adjust, and treat that decision-making items list carries out dispensing decision-making to next using the dispensing decision-making subsystem after adjustment, thus The decision accuracy for launching decision-making next time can be improved.It can be seen that it is a study that the commodity that the application provides, which launch decision system, System, can constantly self-optimization, so as to improve the decision accuracy of system.

When it is implemented, described adaptively adjusted according to first gap, it may include following aspect：1) adjustment institute The decision rule for launching decision-making subsystem is stated, for example, newly-increased, renewal or deletion decision rule etc.；2) the dispensing decision-making is adjusted The boundary condition of subsystem, for example, increasing, modifying or deleting boundary condition etc.；3) pact for launching decision-making subsystem is adjusted Beam condition, for example, newly-increased, renewal or deletion constraint condition etc.；4) processing parameter for launching decision-making subsystem, example are adjusted Such as, parameter value of some parameter etc. is changed；5) tune for launching decision model is carried out according to the actual effect data of launching It is whole.

The staining effect subsystem 103, for obtaining the actual dispensing effect data of the commodity that can be launched, and will The prediction of the actual dispensing effect data of the commodity launched and the commodity launched is launched effect data and compared Compared with acquisition first gap；And first gap is sent out.

The commodity launched refer to, can actually be launched in commodity exhibition by what the dispensing decision-making subsystem determined Show the commodity in the page.The reality is launched effect data and referred to, the true dispensing effect data for the commodity launched, for example, The actual commodity for participating in poly- business to one's profit launch caused actual sales revenue etc. in the cycle at one by the merchandise display page；Institute State prediction dispensing effect data to refer to, pass through the dispensing effect for the commodity that decision-making is treated described in the dispensing decision-making subsystem acquisition Predicted value, for example, supporting the commodity of poly- auction business to one's profit to launch hole position of the commodity examination ＆ verification model in decision system to commodity (resource-niche) sale output is predicted obtained prediction sales volume etc..

It is described to launch staining effect subsystem 103 in the commodity that the application provides launch decision system in indispensable Status, the subsystem to launch decision-making subsystem 101 provide system optimization foundation.The dispensing He of decision-making subsystem 101 A closed loop is formed between the dispensing staining effect subsystem 103, the closed loop causes determining for the dispensing decision-making subsystem 101 Plan precision is continuously available lifting.

Fig. 2 is refer to, it launches the first specific schematic diagram of the embodiment of decision system for the commodity of the application.The business Product, which launch decision system, also to be included：Category plans subsystem 201.

The category plans subsystem 201, and the category for receiving for specific dispensing effect target plans request, and root According to the characteristic of each commodity classification to be planned, the specific dispensing effect target and the category planning mould previously generated Type, generate category planning forecast result；And the category planning forecast result is sent out；And receive the dispensing The category that staining effect subsystem 103 provides plans second gap of the actual result relative to category planning forecast result, and root Adaptively adjusted according to second gap, and carry out category next time using the category planning subsystem 201 after adjustment and advise Draw.

The dispensing effect target includes but is not limited to：The dispensing decision business of commodity is in some period (launching the cycle) Overall operation target, can also be other dispensing effect targets, for example, an average pin for launching all resource-niches in the cycle Sell output etc..

The category planning refers to, to reach commodity classification and the corresponding resource digit that the dispensing effect target is made The physical planning of amount.The category planning actual result refers to, the dispensing of commodity is carried out according to the category planning forecast result Decision-making and the actual category program results for according to the result of decision is launched formed after service implementation.

The commodity classification to be planned has the characteristic of multiple dimensions, and the characteristic of the commodity classification includes Following any characteristic it is at least one：The characteristic of history conclusion of the business effect, the characteristic of history buyer's behavior.

The characteristic of the history conclusion of the business effect, mainly summed up from the daily record of record commodity transaction information anti- The characteristic of the history conclusion of the business effect of commodity classification is reflected, for example, the total business volume of commodity classification, commodity classification is in a throwing Put the consumption sum of average single resource-niche in cycle etc..

The characteristic of buyer's behavior, the reflection business mainly summed up from the daily record of record buyer's behavioral data The characteristic of category purpose buyer's behavior, for example, browsing commodity the class number of users of commodity, purchase business now in one month in the past Category number of users of commodity etc. now.

In addition, the characteristic of commodity classification may also include：Reflect the data of commodity classification unique characteristics, for example, commodity The data such as the commodity amount that classification includes, the Seller Number that commodity classification is related to.

For example, when the category to poly- website to one's profit is planned, the characteristic of commodity classification may include：The classification bag The commodity amount included, the Seller Number that the classification is related to, the users of the classification commodity was browsed in past 3 days, one week, one month The number of users of the classification commodity is counted, bought, the classification commodity are added into the number of users of shopping cart, collect the user of the classification commodity Number, such purpose total business volume, the classification are cheated (i.e. in the poly- average list calculated：Resource-niche) output；In addition, commodity classification Characteristic may also include：The classification is in the conclusion of the business accounting of whole market, and the various data in one month same period refer in the past The data such as mark.

The characteristic of the commodity classification, it can be extracted and obtained from commodity related data by data mining subsystem, example Such as, the characteristic of the history conclusion of the business effect of the commodity classification is excavated from the daily record of record commodity transaction information, from note Record in the daily record of buyer's behavioral data and excavate characteristic of buyer's behavior of the commodity classification etc..

The category plans subsystem 201 after receiving for the category planning request of specific dispensing effect target, By the characteristic according to each commodity classification to be planned and the specific dispensing effect target, pass through the category of the subsystem Plan model, the prediction result of generation category planning.The commodity classification that the prediction result of the category planning includes is described treats The subset of the commodity classification collection of planning.Dispensing decision-making is carried out to commodity according to the prediction result that category is planned, can finally be reached The specific dispensing effect target.

The category planning subsystem 201 is realized is converted into specific product by the overall planning index of specific dispensing effect target The function of class program results.Still the category is planned so that the commodity of poly- auction business to one's profit launch decision system as an example below Subsystem 201 illustrates.

The category planning of poly- auction business to one's profit refers to, to the commodity classification on the merchandise display page and corresponding resource digit Amount is planned.In poly- auction business to one's profit, the category planning subsystem 201 is being received for specific dispensing effect mesh Mark is (such as：The data such as the gross sales amount of the average sale of each resource-niche or all resource-niches) category planning request after, need The characteristic of each commodity classification to be planned is obtained, and according to these characteristics and specific dispensing effect target under The hole position classification and quantity of the poly- activity to one's profit in one dispensing cycle are predicted, for example, specific dispensing effect target is each The average sale of resource-niche is 5,000,000 RMB, and poly- draw can be participated in by planning that subsystem 201 can be predicted by the category Calculate which the commodity classification belonging to lipuid goods includes, and the hole bit quantity of each commodity classification, i.e.,：One commodity class is now The poly- commodity amount for calculating activity can be participated in.During service operation, if the hole bit quantity for predicting some commodity classification is 30, and the commodity class only has 20 commodity that there is sales volume or sales volume to reach a fixed number now after actual poly- activity to one's profit Amount, then category planning actual result is 20, and category planning forecast result is 30.

When the commodity that the application provides, which launch decision system, also includes category planning subsystem 201, the dispensing decision-making Subsystem 101, it is additionally operable to receive the category planning forecast result that the category planning subsystem 201 provides, and is passing through After effect data is launched in each prediction for treating decision-making commodity of dispensing decision model generation, according to the category planning forecast As a result the commodity that can be launched described in determining, the category planning forecast result is met to launch the result of decision.

The commodity that the application provides launch decision system, emphasize it is according to the feedback of dispensing staining effect subsystem 103, Gap between category planning actual result and category planning forecast result, category planning subsystem 201 is carried out from Adjustment is adapted to, and category planning next time is carried out using the category planning subsystem 201 after adjustment, can thus be gradually stepped up The precision of category planning, and then improve the decision accuracy that the commodity launch decision system.

Fig. 3 is refer to, it launches the second specific schematic diagram of the embodiment of decision system for the commodity of the application.The business Product, which launch decision system, also to be included launching effect target prediction subsystem 202.

The dispensing effect target prediction subsystem 202, for receiving the dispensing effect target for the specific dispensing cycle Predictions request, and according to the specific characteristic for launching the cycle, pass through the dispensing effect target prediction model previously generated The dispensing effect target in specific dispensing cycle is predicted, i.e.,：Effect prediction target is launched in generation；And by the dispensing Effect prediction target is sent out；And receive the dispensing effect realistic objective launched staining effect subsystem 103 and provided Relative to the 4th gap of the dispensing effect prediction target, and adaptively adjusted, and used according to the 4th gap Dispensing effect target prediction subsystem after adjustment carries out target prediction next time.

Accordingly, the dispensing staining effect subsystem 103, it is additionally operable to obtain the dispensing effect realistic objective, and will The dispensing effect realistic objective obtains the 4th gap compared with the dispensing effect prediction target；And by institute The 4th gap is stated to be sent out.Wherein, the dispensing effect realistic objective refers to, the actual dispensing effect in a dispensing cycle.

The dispensing effect target includes but is not limited to：The commodity launch decision system will in some specific dispensing cycle The overall operation target reached, can also be other dispensing effect targets, for example, one is launched the flat of all resource-niches in the cycle Sell output etc..The sales volume of commodity is generally related to the factor such as sales season, market, therefore, the operation of different time sections Target is typically different.

For example, the dispensing effect target of poly- lower pair 11 period of auction business to one's profit is (i.e.：Operation target) compare other times The dispensing effect target of section is generally higher by a lot, the dispensing effect mesh of the dispensing effect target of early stage festivals or holidays than the daily period Mark is generally high.In actual applications, effect is launched by launching each history the commodity in cycle to analyze, can be with The variation tendency of operation target is got, these variation tendencies can also have an impact to the prediction for launching effect target, for example, together The percentage of sales volume average annual growth in one period etc..

In embodiment, first by machine learning algorithm, according to the characteristic in each history dispensing cycle and dispensing Effect target prediction model is launched in the training sample set that corresponding record between effect realistic objective is formed, generation；Then, then According to the specific characteristic for launching the cycle and the dispensing effect target prediction model previously generated, to the specific dispensing cycle Dispensing effect target be predicted.

The commodity that the application provides launch decision system, also emphasize and are fed back according to the dispensing staining effect subsystem 103 , it is described launch effect realistic objective and it is described launch effect prediction target between gap, to the dispensing effect target prediction Subsystem 202 is adaptively adjusted, and target is pre- next time using the dispensing effect target prediction subsystem progress after adjustment Survey, can thus step up the precision of prediction for launching effect target, and then improve the commodity and launch determining for decision system Plan precision.

Fig. 4 is refer to, it is the specific schematic diagram of the data mining subsystem of the embodiment of the decision system of the application.Institute Stating commodity dispensing decision system also includes：Data mining subsystem 105, for excavating various features number from commodity related data According to for example, the characteristic of commodity, the characteristic of commodity classification, the characteristic in dispensing cycle etc..

When it is implemented, the data mining subsystem 105 includes the first number only being had an impact to specific decision-making scene According to the second data mining subsystem 1052 for excavating subsystem 1051 and being had an impact to multiple decision-making scenes.

First data mining subsystem 1051, for described in the extraction from the commodity related data of specific decision-making scene First data.

First data refer to, only carry out the data of decision-making for the decision-making subsystem of specific decision-making scene, i.e.,：Scene Special data, for example, " businessman participates in the poly- ability for calculating activity " this characteristic is only used for supporting poly- auction industry to one's profit The dispensing decision-making subsystem of business, and it is not used to other decision-making subsystems；" commodity classes commodity now were browsed in past one month This characteristic of number of users " is only used for category planning subsystem, and is not used to other decision-making subsystems.The specific decision-making The commodity related data of scene includes reflecting the data of specific decision-making scene service present situation and needed by specific decision-making scene Decision data is treated in decision-making subsystem progress decision-making.

Data mining subsystem 105 generally includes multiple first data mining subsystems 1051, each different institute The first data mining subsystem 1051 is stated respectively for different decision-making subsystems, for example, launching the field that decision-making subsystem needs Scape specific features data source is in the first data mining subsystem towards poly- auction business scene to one's profit, category planning subsystem The scene specific features data source needed is in the first data mining subsystem towards category planning scene.Specific decision-making scene Under first data mining subsystem and the scene decision-making subsystem between be usually man-to-man relation.

Second data mining subsystem 1052, for extracting the second data from the commodity related data of enterprise-level.

Second data refer to, the data for being available for multiple decision-making scenes to use, i.e.,：The general data of different scenes, example Such as, this characteristic of commercial quality can be used to support the commodity of poly- auction business to one's profit to launch decision-making subsystem, also can use In the decision-making subsystem of other decision-making scenes, such as category planning subsystem, decision-making subsystem, model evaluation subsystem are invited.

The commodity related data of the enterprise-level includes complete data complete or collected works, the specific decision-making scene under corporate environment The commodity related data of (as launched decision-making scene) is a subset of the commodity related data of the enterprise-level, a subset number According to the decision process that can completely support under some decision-making scene.

The commodity that the application provides launch decision system, emphasize the data that the data mining subsystem 105 provides Two classes can be divided into：The general data of the special data of special scenes, different scenes.Pass through second data mining subsystem 1052 offers are available for the conventional data that multiple decision-making scenes use, thus can be when extending the system, and only increase is with increasing newly First data mining subsystem under decision-making scene corresponding to decision-making subsystem, newly-increased decision-making subsystem can be with other existing decision-making System shares the conventional data that second data mining subsystem 1052 provides, and is extended to so as to effectively reduce system This.

Fig. 5 is refer to, it launches the 3rd specific schematic diagram of the embodiment of decision system for the commodity of the application.The business Product, which launch decision system, also includes operation Monitor And Control Subsystem 203.

The operation Monitor And Control Subsystem 203, the whole decision process of decision system is launched for monitoring the commodity.

When the commodity that the application provides, which launch decision system, occurs any abnormal in decision process, the operation Monitor And Control Subsystem 203 will record the various exceptions captured；The exception includes the exception that subsystems are dished out；The exception It is storable in journal file or database, is used so that system operation attendant analyzes.

Fig. 6 is refer to, it launches the 4th specific schematic diagram of the embodiment of decision system for the commodity of the application.The business Product, which launch decision system, also includes extension subsystem 204.

The extension subsystem 204, throws for newly-increased decision-making subsystem or predicting subsystem to be increased into the commodity Put decision system.

The commodity that the application provides launch decision system, emphasize by the extension subsystem 204 to whole decision-making System is extended, and the extension subsystem 204 is realized increases to institute by newly-increased decision-making subsystem, newly-increased predicting subsystem State the function that commodity launch decision system.It is extended by the extension subsystem 204 to launching decision system, can be effective Reduction system extends cost.

Fig. 7 is refer to, it launches the 5th specific schematic diagram of the embodiment of decision system for the commodity of the application.The business Product, which launch decision system, also includes model evaluation subsystem 205.

The model evaluation subsystem 205, business is launched for receiving the assessment request for particular model, and according to actual Product and corresponding actual dispensing effect data are assessed the particular model, generate forecast assessment result；And will The forecast assessment result is sent out；And receive the actual assessment result phase launched staining effect subsystem and provided Adaptively adjusted for the 5th gap of the forecast assessment result, and according to the 5th gap, and after use adjustment Model evaluation subsystem next particular model is assessed.

The particular model include the particular prediction model that is obtained by machine learning algorithm from training data learning or Specific decision model, for example, launching the dispensing decision model of decision-making subsystem application, the category rule of category planning subsystem application Draw model etc..

With the continuous accumulation of commodity related data, constantly increasing available for the training data for generating the particular model Add；With the continuous development of machine learning techniques, the machine learning algorithm for generating particular model institute foundation is also constantly changing Enter；These factors will result in the need for regenerating the particular model, with the degree of accuracy of lift scheme.The effect of new model is such as What, the precision difference of new model and old model how, be required to be assessed by the model evaluation subsystem 205, with true Fixed actual applicable model, so as to preferably support the whole operation for launching decision system.

Accordingly, the staining effect subsystem 103, is additionally operable to obtain the actual assessment result, and by the reality Assessment result obtains the 5th gap compared with the forecast assessment result；And by the 5th gap to outgoing Send.

Model evaluation technology is a kind of more ripe technology, can be divided into the side assessed before model is employed it Formula and the mode assessed after model is actually applied it.The side assessed before model is employed it Formula is can not to carry out quality evaluation to it before model implementation by observing practical business effect come scoring model, which； The assessment mode that carried out after model is actually applied to it is according to the history commodity pair with actual implementation result data What model was assessed, which can carry out quality evaluation before model implementation to it.In actual applications, can be according to specific Demand, choose any of the above-described kind of assessment mode and model is assessed.

The commodity that the application provides launch decision system, emphasize being fed back according to the model evaluation subsystem 205, described Gap between actual assessment result and the forecast assessment result, the model evaluation subsystem 205 is adaptively adjusted, And next particular model is assessed using the model evaluation subsystem 205 after adjustment, so ensure that model Precision of prediction, and then improve the decision accuracy that the commodity launch decision system.

Fig. 8 is refer to, it launches the 6th specific schematic diagram of the embodiment of decision system for the commodity of the application.In this reality Apply in example, the commodity are launched decision system and specifically included：Launch decision-making subsystem 101 and launch first corresponding to decision-making scene Data mining subsystem 1051-1, category planning subsystem 201 and the first data mining subsystem corresponding to category planning scene 1051-2, launch the first data mining subsystem 1051- corresponding to effect target prediction subsystem 202 and target prediction scene 3rd, the first data mining subsystem 1051-4 corresponding to model evaluation subsystem 205 and model evaluation scene, launch staining effect Subsystem 103, second feature data knowledge excavate subsystem 1052.For the system above-mentioned each part specifically Annexation between bright and each several part, the above-mentioned associated description to various pieces is referred to, here is omitted.

As seen from Figure 8, the commodity, which launch decision system, includes multiple decision-making subsystems, specially launches effect target prediction Subsystem 202, category planning subsystem 201, launch decision-making subsystem 101 and model evaluation subsystem 205, different decision-making subsystems System provides the result of decision of different aspect, and the dispensing result of decision is ultimately formed by more sub- decision loop sections；This processing mode, So that it is configured between entirely launching multiple coherent the scene decision-making subsystems and staining effect subsystem under decision business One big closed loop, when this big closed loop circulation is got up, the dispensing decision business of commodity can reach the operational effect of wisdom.

It should be noted that the commodity shown in Fig. 8, which launch decision system, is carrying out the characteristic of dispensing decision-making when institute foundation According to the characteristic including Knowledge, for example, it is to know that commercial quality, businessman, which participate in the characteristics such as the poly- ability for calculating activity, Know the characteristic of level.The level of knowledge is divided into from the bottom up：Initial data, indication information, knowledge and wisdom.From commodity The process of extraction Knowledge characteristic is in related data：Initial data according to business scenario, which calculates, obtains indication information, More abstract structure information is extracted from indication information, i.e.,：Knowledge.It can be seen that the characteristic of Knowledge be typically based on it is substantial amounts of general Logical indication information calculates generation, for example, according to the common indication information of each dimension and the Knowledge characteristic previously generated Forecast model, generate the characteristic of Knowledge.The characteristic of Knowledge is to realize that " wisdom " launches the basis of decision-making.

In addition, in the actual implementation process of system, effect target, the raising efficiency of decision-making are launched to reach, the application is real Applying the commodity dispensing decision system of example offer also includes：Invite decision-making subsystem 206.

The invitation decision-making subsystem 206, the category planning provided for receiving the category planning subsystem 201 Prediction result, receive and invite the first data mining subsystem 1051-5 and the second feature data knowledge corresponding to decision-making scene The characteristic for the commodity for meeting the category planning forecast result that subsystem 1052 provides is excavated, and according to acquisition Meet the characteristic of the commodity of the category planning forecast result and the invitation decision model previously generated, meet institute to described The dispensing effect for stating the commodity of category planning forecast result is predicted, and is launched effect data according to prediction and determined what can be invited Commodity, formed and invite the result of decision；And receive the business invited for launching staining effect subsystem 103 and providing The actual dispensing effect data of product relative to the commodity invited prediction launch effect data the 3rd gap, and according to 3rd gap is adaptively adjusted, and carries out invitation decision-making to commodity using the invitation decision-making subsystem after adjustment.

The invitation decision-making subsystem 206 realizes following function：From the shiploads of merchandise for meeting category planning forecast result It is that can launch the higher commodity of probability of commodity to filter out by decision-making.For this part commodity filtered out, because it is by decision-making For the probability that can launch is higher, to launch effect target more effective to promoting to reach, therefore, this part commodity can be carried out actively Invite, promote it to participate in the dispensing decision-making of the commodity.By the way that the invitation decision-making subsystem 206 is added in the system, Can further lifting system wisdom operational effect.

Corresponding, the dispensing staining effect subsystem 103, it is additionally operable to the actual throwing of commodity that can be invited described in obtaining Effect data is put, and the prediction of the actual dispensing effect data of the commodity invited and the commodity invited is launched Effect data is compared, and obtains the 3rd gap；And the 3rd gap is sent out.

Below using poly- auction business to one's profit as application background, decision system is launched to the commodity shown in Fig. 8 and made furtherly It is bright.In poly- commodity auction business to one's profit, the basic running that commodity launch decision system comprises the following steps：

1) operation personnel launches decision system transmission to the commodity by its client first and is directed to the specific dispensing cycle (such as：During 2016 double 11) the request of dispensing effect target prediction.

2) after the dispensing effect target prediction subsystem 202 of the commodity dispensing decision system receives the request, The characteristic in dispensing cycle is obtained by the first data mining subsystem 1051-3 corresponding to target prediction scene first (such as：Double 11, market environment such as is clearly better at the characteristic for upper 1 year), and according to these characteristics, by pre- Dispensing effect target of the dispensing effect target prediction model first generated to the dispensing cycle is (such as：The average pin of all resource-niches Sell volume) it is predicted, generation launches effect prediction target (such as：1,000 ten thousand) average sale of all resource-niches is；Then, then The category planning request for the target is formed, and sends this request to the category planning subsystem 201.

3) the category planning subsystem 201 plans field by category first after above-mentioned category planning request is received First data mining subsystem 1051-2 corresponding to scape and/or the second feature data knowledge excavate subsystem 1052, obtain Each commodity classification to be planned is (such as：Include 100 commodity classifications to be planned altogether) characteristic, and according to these features Data and above-mentioned dispensing effect prediction target, by the category plan model of the subsystem, the poly- of above-mentioned dispensing cycle is calculated The hole position classification and quantity of activity are predicted, and generation category planning forecast result is (such as：It is determined that wherein 30 commodity classifications have Resource-niche)；Then, then by the category planning forecast result send to the dispensing decision-making subsystem 101.

4) commodity for treating decision-making launched decision-making subsystem 101 and sent in the client for receiving operation personnel List is (such as：Including 10,000 commodity for participating in poly- to one's profit auction) decision requests after, first by launching decision-making scene Corresponding first data mining subsystem 1051-1 and/or the second feature data knowledge excavate subsystem 1052, obtain each The characteristic of individual commodity, and according to the characteristic of each commodity, above-mentioned category planning forecast result and the throwing previously generated Put decision model, determine each commodity class now launch commodity；Finally, the commodity of launching of each commodity class now are shown Show in the merchandise display page of poly- platform to one's profit, the page is that each commodity classification is set according to above-mentioned category planning forecast result Corresponding resource-niche is put, each can launch commodity and be arranged in its resource-niche of affiliated commodity class now.

Commodity in the items list for treating decision-making, it can be obtained by the invitation decision-making subsystem 206.It is described to invite Please decision-making subsystem 206 receive the category planning forecast result that category planning subsystem 201 provides, and pass through invitation First data mining subsystem 1051-5 corresponding to decision-making scene and the second feature data knowledge excavate subsystem 1052, obtain The characteristic for the commodity for meeting the category planning forecast result is taken, and meets that the category is planned in advance according to the described of acquisition The invitation decision model surveyed the characteristic of the commodity of result and previously generated, meets the category planning forecast result to described The dispensing effects of commodity be predicted, and the commodity that effect data determines to invite are launched according to prediction；Operation personnel will be with The seller of these commodity is linked up, and it is that these commodity participate in poly- auction activity to one's profit to invite seller.

In poly- commodity auction business to one's profit, the model evaluation subsystem 205 that commodity launch decision system is receiving pin To particular model (such as：Launch effect target prediction model, category plan model, dispensing decision model or invite decision model) After assessment request, first by the first data mining subsystem 1051-4 corresponding to model evaluation scene, actual dispensing business is obtained Product and corresponding actual dispensing effect data, and launch commodity and corresponding actual dispensing effect data according to actual The particular model assessed request is assessed；Then, it can be determined whether to substitute using the particular model according to assessment result and worked as The preceding model used.

In poly- to one's profit commodity auction business, the dispensing staining effect subsystem 103 that commodity launch decision system can be towards Other subsystems provide carries out adaptive adjustment gap data used for subsystems respectively, for example, to the throwing Put decision-making subsystem 101 provide can launch commodity actual dispensing effect data and prediction launch effect data between gap, To category planning subsystem 201 provide category planning actual result relative to the gap of category planning forecast result, to institute State and difference of the offer of the effect target prediction subsystem 202 dispensing effect realistic objective relative to the dispensing effect prediction target is provided Away from etc..

In addition, in poly- commodity auction business to one's profit, commodity launch the operation Monitor And Control Subsystem 203 of decision system, will supervise The whole service process that commodity launch decision system is controlled, records the various abnormal conditions captured, so that operation personnel's reference makes With.

In summary, decision system is launched by the commodity provided in poly- commodity auction business to one's profit using the application, So that multiple coherent scene decision-making subsystems under whole poly- to one's profit commodity auction business and launch staining effect subsystem it Between be configured to a big closed loop, poly- to one's profit commodity auction business reaches the operational effect of wisdom.

Corresponding with above-mentioned commodity dispensing decision system, the application also provides a kind of commodity and launches decision-making technique.It please join Fig. 9 is examined, it is the flow chart for the embodiment that a kind of commodity that the application provides launch decision-making technique, and the present embodiment and first are implemented Example content identical part repeats no more, and refers to the appropriate section in embodiment one.A kind of commodity that the application provides are launched Decision-making technique comprises the following steps：

Step S101：Obtain the items list for treating decision-making.

The items list for treating decision-making includes at least one commodity launched in the merchandise display page for treating decision-making, For example, treat that the items list of decision-making includes the multiple commodity for participating in poly- auction business to one's profit, by supporting poly- auction to one's profit The commodity of business launch decision system (i.e.：The current commodity launches decision system), it can be filtered out from registration commodity The poly- commodity for calculating activity are participated in qualification, and commodity dispensing decision system, which is responsible for providing, launches the result of decision, i.e.,：Which commodity determined Can be by auditing, can participate in poly- activity to one's profit.

Step S102：Decision system is launched by current commodity dispensing decision-making is carried out to each commodity for treating decision-making, obtained Launch the result of decision.

Described current commodity launches the commodity that decision system refers to be being currently used and launches decision system.This step makes The word of decision system one is launched with current commodity, wherein it is continuous in the present embodiment that " current ", which illustrates that commodity launch decision system, Amendment, developing, commodity that this step uses launch decision system is currently valid dispensing decision system, but when crossing one section Between, the dispensing decision system just may be varied from due to the actual difference for launching effect and prediction dispensing effect.It is real The decision accuracy deficiency that effect launches the Discrepancy Description dispensing decision system of effect with prediction is launched on border, therefore, it is necessary to current Launch decision system to be adjusted so that prediction launches effect and launches effect as close possible to actual.

Each commodity that to launch decision system by the current commodity includes to the items list for treating decision-making enter Row launches decision-making, it is necessary first to the various dimensions characteristic of each commodity is obtained, then further according to the various dimensions characteristic of commodity It is predicted according to the dispensing effect to each commodity, and effect data is launched according to prediction and determines whether the commodity can be launched, from And formed and launch the result of decision.

Described current commodity launches decision system to be included in the characteristic for carrying out dispensing decision-making when institute foundation：By knowledge The characteristic for the Knowledge that subsystem extracts from commodity related data is excavated, for example, commercial quality, businessman participate in dispensing Activity is (such as：It is poly- to calculate activity) the characteristic such as ability i.e. to be extracted by knowledge excavation subsystem from commodity related data Knowledge characteristic.The characteristic of Knowledge is to realize that " wisdom " launches the basis of decision-making.

The characteristic that the current commodity launches decision system progress decision-making institute foundation can be divided into only to specific decision-making Fisrt feature data that scene has an impact and the second feature data that multiple decision-making scenes are had an impact.

The fisrt feature data refer to, only carry out the characteristic of decision-making for the decision-making subsystem of specific decision-making scene According to that is,：The special characteristic of scene, for example, this characteristic of the ability of the poly- activity to one's profit of businessman's participation is only used for supporting The commodity of poly- auction business to one's profit launch decision-making subsystem, and are not used to other decision-making subsystems.

The second feature data refer to, the characteristic for being available for multiple decision-making scenes to use, i.e.,：Different scenes are general Characteristic, for example, this characteristic of commercial quality can be used to support the dispensing decision-making subsystem of poly- auction business to one's profit, It can be additionally used in the decision-making subsystem of other decision-making scenes, such as category planning subsystem, invite decision-making subsystem, model evaluation subsystem System etc..

In the present embodiment, data mining subsystem includes the excavation subsystem of fisrt feature data, second feature data Excavation subsystem.Wherein, the excavation subsystem of the fisrt feature data, for the commodity dependency number from specific decision-making scene According to the middle extraction fisrt feature data；The excavation subsystem of the second feature data is related for the commodity from enterprise-level Second feature data described in extracting data.

For example, businessman participates in, to gather this characteristic of ability to one's profit be by launching decision-making scene towards poly- commodity to one's profit The fisrt feature data that first data mining subsystem provides, this characteristic of commercial quality are by towards group of Alibaba The second feature data that provide of the second data mining subsystem, commercial quality characteristic can be used to poly- to one's profit commodity and launches Decision-making scene, it may also be used for other decision-making scenes, such as category plan scene.

It can be that the integrated decision-making system formed is combined by multiple decision-making subsystems that described current commodity, which launches decision system, System.Different decision-making subsystems provide the result of decision of different aspect, are ultimately formed by more sub- decision loop sections and launch decision-making knot Fruit, it thus can reach the operational effect of wisdom.

This step S102 can be realized in the following way：According to the category planning forecast result previously generated, treated to each The commodity of decision-making carry out dispensing decision-making.The category planning refers to, to reach the commodity class that specific dispensing effect target is made The real needs planning of mesh and corresponding resource bit quantity, for example, in poly- auction business to one's profit, category program results includes joining Classification affiliated to lipuid goods, each commodity class cheat bit quantity etc. and run related parameter now.

The dispensing effect target, can be totality of the dispensing decision business in some period (launching the cycle) of commodity Target is runed, for example, poly- auction business to one's profit can be set in the operation target in certain stage as 50,000,000 RMB；It is described to launch effect Fruit target, can also be other dispensing effect targets, for example, an average sale output for launching all resource-niches in the cycle Deng.

The category planning forecast result refers to that the category that decision system offer is launched by the current commodity plans knot Fruit.The category planning actual result refers to, dispensing decision-making and the basis of commodity are carried out according to the category planning forecast result Launch the actual category program results that the result of decision formed after service implementation.

In the specific implementation, the category planning forecast result, following steps can be used to generate：1) receive and be directed to specific throwing Put the category planning request of effect target；2) characteristic of each commodity classification to be planned is obtained；3) according to described specific Effect target, the characteristic of the commodity classification and the category plan model previously generated are launched, generates category planning forecast As a result.

In the specific implementation, the specific dispensing effect target, can use following steps to generate：1) receive and be directed to specific throwing Put the dispensing effect target prediction request in cycle；2) the specific characteristic for launching the cycle is obtained；3) according to described specific The predicted value of effect target is launched in the dispensing effect target prediction model launched the characteristic in cycle and previously generated, generation, As the specific dispensing effect target.

The commodity that the embodiment of the present application provides launch decision-making technique, and the items list of plan only not co-pending, which provides, launches decision-making As a result, in addition to launching the prediction of effect target and planning to category, and according to various predicted values and corresponding Gap between actual value carries out the process of decision system adjustment；This processing mode so that whole to launch under decision business A big closed loop is configured between multiple coherent decision-making scenes and staining effect system, when this big closed loop circulation is got up When, whole system can reach the operational effect of wisdom.

Step S103：Obtain the actual dispensing effect data after commodity are launched according to the dispensing result of decision.

Described current commodity launches dispensing effect prediction data of the decision system provided in whole decision process can To be multi-angle, many dispensing effect datas.Therefore, described dispensing effect includes many-sided dispensing effect.

Described actual dispensing effect data is storable in database, in the file of journal file or other forms. In the present embodiment, actual effect data of launching is stored in database, and the actual dispensing can be obtained by data base querying Effect data.

Step S105：By the actual effect data of launching with generating the described current of dispensing result of decision institute foundation Commodity, which launch the prediction that decision system provides and launch effect data, to be compared, obtain it is described it is actual launch effect data relative to The gap of effect data is launched in the prediction.

This step by the actual dispensing effect data that previous step is got with generate it is described dispensing the result of decision when pass through The current commodity is launched the prediction dispensing effect data that decision system provides and is compared, to obtain the actual dispensing effect Data launch the gap of effect data relative to the prediction.

In the specific implementation, effect is launched if there is many-side, then needs to launch each side into the reality of effect Launch effect data and the corresponding prediction is launched effect data and is compared respectively, to obtain various dispensing effects pair The gap answered.

Step S107：Commodity are carried out according to the gap and launch decision system adjustment.

After getting the gap between actual dispensing effect data and prediction dispensing effect data, you can according to the gap pair The dispensing decision-making of the commodity is adjusted.

In the specific implementation, effect is launched if there is many-side, then needs to consider the adjustment decision-making of each side gap System.Described carries out commodity dispensing decision system adjustment according to the gap, and including but not limited to following method of adjustment is extremely Few one：Adjust commodity and launch the decision rule of decision system, boundary condition, the adjustment commodity of adjustment commodity dispensing decision system Launch the constraints of decision system, adjustment commodity launch the processing parameter of decision system, according to newly-increased actual dispensing effect Data carry out launching adjustment of decision model etc..By the way that commodity are launched with the adjustment of decision system, itself and handled number can be made According to Statistical Distribution Characteristics, architectural feature be adapted, to obtain optimal decision-making results.

The core that described commodity launch decision system is to launch decision model, can be predicted by launching decision model Data.Decision system is adjusted including the adjustment to launching decision model.

Step S109：After the completion of the commodity launch decision system adjustment, the commodity of the renewal of acquisition are launched into decision-making system System launches decision system as current commodity.

After previous step completes decision system adjustment, the decision system obtained compares the decision system originally used, The decision system exactly updated；The decision system of the renewal can substitute original decision system at once as current decision system System.When performing step S103, it is exactly the decision system by this renewal that the current commodity, which launches decision system,.

Using this method, it can carry out launching decision-making while carrying out system update, and thrown at once after system update Enter to use, then to treat that the items list of decision-making provides the result of decision under new decision system, and go round and begin again, realize actual Launching the mutual forward direction of effect data accumulation and decision system improves circulation, rapid to improve decision system quality, is lifted Decision accuracy.

In the specific implementation, the method that the application provides also includes：According to the gap, adjust the current commodity and launch Decision system carries out the characteristic of decision-making institute foundation and the current commodity launches the decision making algorithm of decision system use extremely Few one.In the specific implementation, the step is typically to be completed by system developer.

In summary, the above-mentioned process being adjusted according to the gap to decision system can be referred to as to oneself of unmanned participation Adjustment process is adapted to, the current commodity dispensing decision system progress decision-making institute foundation is adjusted according to the gap by described Characteristic and the current commodity launch the process of at least one of the decision making algorithm that decision system uses, and are referred to as having ginseng With total system iterative process.

In the above-described embodiment, there is provided a kind of commodity launch decision-making technique, and corresponding, the application also provides A kind of commodity launch decision making device.The device is corresponding with the embodiment of the above method.

Figure 10 is refer to, it launches the schematic diagram of decision making device embodiment for the commodity of the application.Due to device embodiment Embodiment of the method is substantially similar to, so describing fairly simple, the relevent part can refer to the partial explaination of embodiments of method. Device embodiment described below is only schematical.

A kind of commodity of the present embodiment launch decision making device, including：Items list acquiring unit 101, it is co-pending for obtaining The items list of plan；Decision package 102 is launched, each commodity for treating decision-making are entered for launching decision system by current commodity Row launches decision-making, obtains and launches the result of decision；It is actual to launch effect data acquiring unit 103, for obtaining according to the dispensing The result of decision launches the actual dispensing effect data after commodity；Contrast on effect unit 104, for the reality to be launched into effect number According to the prediction dispensing effect number launched decision system with generating the current commodity for launching result of decision institute foundation and provided According to being compared, the actual gap launched effect data and effect data is launched relative to the prediction is obtained；System call interception Unit 105, decision system adjustment is launched for carrying out commodity according to the gap；System update unit 106, for the commodity After the completion of launching decision system adjustment, the commodity of the renewal of acquisition are launched into decision system and launch decision-making system as current commodity System.

Figure 11 is refer to, it is the schematic diagram of the electronic equipment embodiment of the application.Because apparatus embodiments are substantially similar In embodiment of the method, so describing fairly simple, the relevent part can refer to the partial explaination of embodiments of method.It is described below Apparatus embodiments it is only schematical.

The a kind of electronic equipment of the present embodiment, the electronic equipment include：Processor 101；And memory 103, for depositing The program that commodity launch decision-making technique is stored up, after the equipment is powered and runs the program that the commodity launch decision-making technique, is performed following Step：Obtain the items list for treating decision-making；Decision system is launched by current commodity to launch each commodity for treating decision-making Decision-making, obtain and launch the result of decision；Obtain the actual dispensing effect data after commodity are launched according to the dispensing result of decision；Will The actual effect data of launching carries with the generation current commodity dispensing decision system for launching result of decision institute foundation The prediction of confession is launched effect data and is compared, and obtains the actual effect data of launching and launches effect number relative to the prediction According to gap；Commodity are carried out according to the gap and launch decision system adjustment；After the completion of the commodity launch decision system adjustment, The commodity of the renewal of acquisition are launched into decision system and launch decision system as current commodity.

The commodity that the application provides launch decision-making technique, by it is actual launch effect data provide with dispensing decision system it is pre- Survey dispensing effect data to be compared, launch decision system according to actual effect data of launching relative to prediction dispensing effect data Gap adaptively adjusted, and after new dispensing decision system is formed, think immediately available as current decision system of launching Next items list for treating decision-making provides the items list result of decision for treating decision-making.The method provided using the application, is being carried While for launching the result of decision, the actual gap launched effect data and effect data is launched relative to prediction of acquisition, and according to This gap adjusts dispensing decision system in real time；This processing mode, decision accuracy can be effectively improved.

Although the application is disclosed as above with preferred embodiment, it is not for limiting the application, any this area skill Art personnel are not being departed from spirit and scope, can make possible variation and modification, therefore the guarantor of the application Shield scope should be defined by the scope that the application claim is defined.

In a typical configuration, computing device includes one or more processors (CPU), input/output interface, net Network interface and internal memory.

Internal memory may include computer-readable medium in volatile memory, random access memory (RAM) and/or The forms such as Nonvolatile memory, such as read-only storage (ROM) or flash memory (flash RAM).Internal memory is computer-readable medium Example.

1st, computer-readable medium can be by any side including permanent and non-permanent, removable and non-removable media Method or technology realize that information stores.Information can be computer-readable instruction, data structure, the module of program or other numbers According to.The example of the storage medium of computer includes, but are not limited to phase transition internal memory (PRAM), static RAM (SRAM), dynamic random access memory (DRAM), other kinds of random access memory (RAM), read-only storage (ROM), Electrically Erasable Read Only Memory (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc are read-only Memory (CD-ROM), digital versatile disc (DVD) or other optical storages, magnetic cassette tape, tape magnetic rigid disk storage or Other magnetic storage apparatus or any other non-transmission medium, the information that can be accessed by a computing device available for storage.According to Herein defines, and computer-readable medium does not include non-temporary computer readable media (transitory media), such as modulates Data-signal and carrier wave.

2nd, it will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer program production Product.Therefore, the application can use the embodiment in terms of complete hardware embodiment, complete software embodiment or combination software and hardware Form.Moreover, the application can use the computer for wherein including computer usable program code in one or more can use The computer program product that storage medium is implemented on (including but is not limited to magnetic disk storage, CD-ROM, optical memory etc.) Form.

Claims (21)

1. a kind of commodity launch decision system, it is characterised in that including：Launch decision-making subsystem and launch staining effect subsystem System；

The dispensing decision-making subsystem, for receiving the decision requests of the items list for treating decision-making, and according to each commodity Characteristic and the dispensing decision model that previously generates, it is determined that the commodity that can be launched；And receive the dispensing staining effect The actual dispensing effect data for the commodity launched that subsystem provides is thrown relative to the prediction of the commodity launched The first gap of effect data is put, and is adaptively adjusted according to first gap, and uses the dispensing decision-making after adjustment Subsystem treats that decision-making items list carries out dispensing decision-making to next；

The dispensing staining effect subsystem, for obtaining the actual dispensing effect data of the commodity that can be launched, and by institute The actual dispensing effect data for the commodity that can be launched is stated compared with the prediction dispensing effect data of the commodity launched, Obtain first gap；And first gap is sent out.

2. commodity according to claim 1 launch decision system, it is characterised in that described to be carried out according to first gap Adaptive adjustment, including：

Adjust the decision rule of the dispensing decision-making subsystem；

Adjust the boundary condition of the dispensing decision-making subsystem；

Adjust the constraints of the dispensing decision-making subsystem；

Adjust the processing parameter of the dispensing decision-making subsystem；

The adjustment for launching decision model is carried out according to the actual effect data of launching.

3. commodity according to claim 1 launch decision system, it is characterised in that also include：Category planning forecast subsystem System；

The category plans subsystem, and the category for receiving for specific dispensing effect target plans request, and according to each The characteristic of commodity classification to be planned, the specific category plan model launched effect target and previously generated, generation Category planning forecast result；And the category planning forecast result is sent out；And the reception dispensing effect is anti- Present the category that subsystem provides and plan second gap of the actual result relative to category planning forecast result, and according to described second Gap is adaptively adjusted, and is carried out category next time using the category planning subsystem after adjustment and planned；Wherein, the product Class planning actual result refers to, caused actual category after the dispensing decision-making of commodity is carried out according to the category planning forecast result Program results；

The dispensing staining effect subsystem, it is additionally operable to obtain the category planning actual result, and the category is planned in fact Border result obtains second gap compared with the category planning forecast result；It is and second gap is outside Send；

The dispensing decision-making subsystem, it is additionally operable to receive the category planning forecast knot that the category planning subsystem provides Fruit, and the commodity that can be launched according to determining the category planning forecast result.

4. commodity according to claim 3 launch decision system, it is characterised in that also include：Invite decision-making subsystem；

The invitation decision-making subsystem, the category planning forecast result provided for receiving the category planning subsystem, And according to the characteristics of the commodity for meeting the category planning forecast result and the invitation decision model previously generated, it is determined that can The commodity of invitation；And receive the actual dispensing effect for launching the commodity invited that staining effect subsystem provides Fruit data launch the 3rd gap of effect data relative to the prediction of the commodity invited, and are entered according to the 3rd gap The adaptive adjustment of row, and invitation decision-making is carried out to commodity using the invitation decision-making subsystem after adjustment；

The dispensing staining effect subsystem, the actual dispensing effect data of commodity that can be invited described in obtaining is additionally operable to, and will The prediction of the actual dispensing effect data of the commodity invited and the commodity invited is launched effect data and compared Compared with acquisition the 3rd gap；And the 3rd gap is sent out.

5. commodity according to claim 3 launch decision system, it is characterised in that also include：Launch effect target prediction Subsystem；

The dispensing effect target prediction subsystem, the dispensing effect target prediction for receiving for the specific dispensing cycle please Ask, and launched according to the specific characteristic for launching the cycle and the dispensing effect target prediction model previously generated, generation Effect prediction target；And the dispensing effect prediction target is sent out；And receive dispensing staining effect The dispensings effect realistic objective that system provides is relative to the 4th gap for launching effect prediction target, and according to the described 4th Gap is adaptively adjusted, and carries out target prediction next time using the dispensing effect target prediction subsystem after adjustment；

The dispensing staining effect subsystem, it is additionally operable to obtain the dispensing effect realistic objective, and the dispensing effect is real Border target obtains the 4th gap compared with the dispensing effect prediction target；It is and the 4th gap is outside Send.

6. the commodity according to claim any one of 1-5 launch decision system, it is characterised in that also include：Model evaluation Subsystem；

The model evaluation subsystem, for receiving the assessment request for particular model, and according to it is actual launch commodity and with It is actual corresponding to it to launch effect data, generate forecast assessment result；And the forecast assessment result is sent out；With And receive and described launch actual assessment result that staining effect subsystem provides relative to the 5th poor of the forecast assessment result Away from, and adaptively adjusted according to the 5th gap, and using the model evaluation subsystem after adjustment to next specific Model is assessed；

The dispensing staining effect subsystem, is additionally operable to obtain the actual assessment result, and by the actual assessment result with The forecast assessment result is compared, and obtains the 5th gap；And the 5th gap is sent out.

7. the commodity according to claim any one of 1-5 launch decision system, it is characterised in that also include：Data mining Subsystem；

The data mining subsystem includes：The first data mining subsystem for only being had an impact to specific decision-making scene and to more The second data mining subsystem that individual decision-making scene has an impact；

First data mining subsystem, for extracting the first data from the commodity related data of specific decision-making scene；

Second data mining subsystem, for extracting the second data from the commodity related data of enterprise-level.

8. the commodity according to claim any one of 1-5 launch decision system, it is characterised in that also include：Operation monitoring Subsystem；

The operation Monitor And Control Subsystem, the decision process of decision system is launched for monitoring the commodity.

9. the commodity according to claim any one of 1-5 launch decision system, it is characterised in that also include：Extend subsystem System；

The extension subsystem, decision-making system is launched for newly-increased decision-making subsystem or predicting subsystem to be increased into the commodity System.

10. a kind of commodity launch decision-making technique, it is characterised in that including：

Obtain the items list for treating decision-making；

Decision system is launched by current commodity each commodity for treating decision-making are carried out with dispensing decision-making, obtained and launch the result of decision；

Obtain the actual dispensing effect data after commodity are launched according to the dispensing result of decision；

The actual effect data of launching is launched into decision-making with the generation current commodity for launching result of decision institute foundation The prediction that system provides is launched effect data and is compared, and obtains the actual effect data of launching and is launched relative to the prediction The gap of effect data；

Commodity are carried out according to the gap and launch decision system adjustment；

After the completion of the commodity launch decision system adjustment, the commodity of the renewal of acquisition are launched into decision system as current commodity Launch decision system.

11. commodity according to claim 10 launch decision-making technique, it is characterised in that the dispensing effect includes many-side Launch effect；

It is described that the reality is launched into effect data compared with the prediction dispensing effect data, in the following way：

Each side is launched into the actual dispensing effect data of effect and effect data point is launched in the corresponding prediction It is not compared；

Accordingly, the gap includes many-sided gap for launching effect；

It is described that commodity dispensing decision system adjustment is carried out according to the gap, in the following way：

The gap that effect is launched according to each side carries out commodity dispensing decision system adjustment.

12. commodity according to claim 10 launch decision-making technique, it is characterised in that described to enter to do business according to the gap Product launch decision system adjustment, include following method of adjustment at least one：

Adjust the decision rule that commodity launch decision system；

Adjust the boundary condition that commodity launch decision system；

Adjust the constraints that commodity launch decision system；

Adjust the processing parameter that commodity launch decision system；

According to newly-increased actual dispensing effect data launch the adjustment of decision model.

13. commodity according to claim 10 launch decision-making technique, it is characterised in that described obtain is determined according to the dispensing Plan result launches the actual dispensing effect data after commodity, in the following way：

By data base querying, the actual dispensing effect data is obtained.

14. commodity according to claim 10 launch decision-making technique, it is characterised in that the current commodity launches decision-making system The characteristic that system carries out decision-making institute foundation includes the fisrt feature data only being had an impact to specific decision-making scene and to multiple The second feature data that decision-making scene has an impact.

15. commodity according to claim 10 launch decision-making technique, it is characterised in that the current commodity launches decision-making system The characteristic of system progress decision-making institute foundation includes the characteristic of Knowledge.

16. commodity according to claim 10 launch decision-making technique, it is characterised in that described launched by current commodity is determined Plan system carries out dispensing decision-making to each commodity for treating decision-making, in the following way：

According to the category planning forecast result previously generated, dispensing decision-making is carried out to each commodity for treating decision-making.

17. commodity according to claim 16 launch decision-making technique, it is characterised in that the category planning forecast result, Generated using following steps：

Receive the category planning request for specific dispensing effect target；

Obtain the characteristic of each commodity classification to be planned；

According to specific dispensing effect target, the characteristic of the commodity classification and the category plan model previously generated, Generate the category planning forecast result.

18. commodity according to claim 17 launch decision-making technique, it is characterised in that the specific dispensing effect target, Generated using following steps：

Receive the dispensing effect target prediction request for the specific dispensing cycle；

Obtain the specific characteristic for launching the cycle；

Effect is launched according to the specific characteristic for launching the cycle and the dispensing effect target prediction model previously generated, generation The predicted value of fruit target, as the specific dispensing effect target.

19. commodity according to claim 10 launch decision-making technique, it is characterised in that also include：

According to the gap, adjust the current commodity and launch the spy that decision system launch the commodity of decision-making institute foundation Sign and the current commodity launch at least one for the decision making algorithm that decision system uses.

20. a kind of commodity launch decision making device, it is characterised in that including：

Items list acquiring unit, the items list of decision-making is treated for obtaining；

Decision package is launched, dispensing decision-making is carried out to each commodity for treating decision-making for launching decision system by current commodity, Obtain and launch the result of decision；

It is actual to launch effect data acquiring unit, for obtaining the actual dispensing after commodity are launched according to the dispensing result of decision Effect data；

Contrast on effect unit, for by described in the actual dispensing effect data and the generation dispensing result of decision institute foundation Current commodity is launched the prediction dispensing effect data that decision system provides and is compared, and obtains the reality and launches effect data phase The gap of effect data is launched for the prediction；

System call interception unit, decision system adjustment is launched for carrying out commodity according to the gap；

System update unit, after the completion of launching decision system adjustment for the commodity, the commodity of the renewal of acquisition are launched and determined Plan system launches decision system as current commodity.

21. a kind of electronic equipment, it is characterised in that including：

Processor；And

Memory, for storing the program for realizing that commodity launch decision-making technique, the equipment, which is powered and runs the commodity, launches decision-making After the program of method, following step is performed：Obtain the items list for treating decision-making；Decision system is launched to each by current commodity Treat that the commodity of decision-making carry out dispensing decision-making, obtain and launch the result of decision；Obtain after launching commodity according to the dispensing result of decision Actual dispensing effect data；The actual effect data of launching is worked as with generating described in dispensing result of decision institute foundation Preceding commodity are launched the prediction dispensing effect data that decision system provides and are compared, and it is relative that the acquisition reality launches effect data The gap of effect data is launched in the prediction；Commodity are carried out according to the gap and launch decision system adjustment；The commodity are thrown After the completion of putting decision system adjustment, the commodity of the renewal of acquisition are launched into decision system and launch decision system as current commodity.