CN108154275A - Automobile residual value prediction model and Forecasting Methodology based on big data - Google Patents
Automobile residual value prediction model and Forecasting Methodology based on big data Download PDFInfo
- Publication number
- CN108154275A CN108154275A CN201711479877.3A CN201711479877A CN108154275A CN 108154275 A CN108154275 A CN 108154275A CN 201711479877 A CN201711479877 A CN 201711479877A CN 108154275 A CN108154275 A CN 108154275A
- Authority
- CN
- China
- Prior art keywords
- prediction
- library
- function
- vehicle
- influence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses automobile residual value prediction model and Forecasting Methodology based on big data, which includes step：S11, the vehicle information and predicted time for obtaining object vehicle input by user；S12, according to vehicle information input by user, with reference to preset model data library, matching obtains corresponding vehicle configuration information；S13, after choosing corresponding function and parameter from the function data library of prediction model, Prediction Parameters library, influence factor prediction library and prediction error time Sequence Trend library, vehicle information and predicted time are substituted into prediction model to calculate, obtain prediction residual value of the object vehicle in predicted time.The present invention can comprehensively carry out automobile residual value prediction with science, improve forecasting accuracy, and precision of prediction is high, can be widely applied in automobile residual value assessment industry.
Description
Technical field
The present invention relates to data processing field, more particularly to the automobile residual value prediction model based on big data and prediction side
Method.
Background technology
With business lease in automobile industry and the rise of Personal Rental, value of leass, core are being formulated in the prediction of automobile residual value
Calculating tenantry venture etc. has key effect, and residual value prediction is increasingly paid attention to.If being predicted without accurate residual value, automobile is rented
Rent to be developed well.There are mainly three types of existing residual value Forecasting Methodologies：First, by the present residual value situation in market+
Artificial experience obtains predicted value.This method excessively relies on manually, it is impossible to judge accuracy, all not have in history residual value especially
In the case of very long trend, it is difficult to ensure the accuracy of prediction.2nd, the present residual value situation in market is superimposed with vehicle age at the end of prediction
Extrapolate to obtain predicted value, such as A vehicles now it is observed that the residual value of 1 year is 0.85 with mileage growth pattern, then 3 years residual values
It predicts and is then：The influence of ten thousand kilometers of the influence -3 in 2 years vehicle ages of 0.85-.On the one hand this method can not be observed what is just listed
New car to residual value situation gives a forecast, and on the other hand, this method only considers the variation of vehicle source oneself factor, do not consider market
It influences.It is likely in the residual value in present market and future market with the residual value of vehicle because external market environment changes and becomes
Change, if this part is ignored, there can be very big risk.Such as certain vehicle just lists, and is in that supply falls short of demand the phase, observation
It is higher to market last issue residual value.If extrapolating according to this method obtains residual value, it will also continue its high residual value after 3 years, merely because vehicle
The increase of age and mileage and given discount.Because the part residual value that supply falls short of demand raises remains in prediction and suffers.But
If no longer there is the situation that supply falls short of demand after 3 years, residual value can be significantly lower than the prediction result that this method obtains.3rd, in side
Certain following trend obtained by machine learning is superimposed on the basis of method two again.This method, obtained residual value prediction result
Lack explanatory, there is no the forecast systems of science, can not ensure accuracy.
Generally speaking, lack science, the automobile residual value Forecasting Methodology of system at present, accurately automobile residual value can not be carried out
Prediction.
Invention content
In order to solve the technical issues of above-mentioned, the object of the present invention is to provide the automobile residual value prediction models based on big data
And Forecasting Methodology.
The technical solution adopted by the present invention to solve the technical problems is：
Automobile residual value prediction model based on big data, the prediction model are established by following steps and obtained：
S01, a large amount of historical trading datas for obtaining automobile in nearest a period of time；
S02, historical trading data is carried out to intersect data cleaning treatment, rejecting does not have representative historical trading number
According to；
S03, the historical trading data after cleaning is divided into the training data in N number of period, with reference to preset vehicle number
According to library, the training data in each period is resolved into price and the multiple and relevant influence factor of used car price；N is
Natural number；
S04, the analysis result according to all historical trading datas, for each vehicle, from preset alternative function library
Choose suitable anticipation function and the influence function of each influence factor, component function database；
S05, the analysis result based on the training data in each period are calculated respectively using optimization algorithm so that predicting
The function parameter combination of the cost function of function and each influence function minimum, so as to establish history parameters corresponding with the period
Library；
S06, for each vehicle, according to its different history parameters library within N number of period, prediction obtains different parameters
Change with time trend, establishes Prediction Parameters library；
S07, prediction obtain the trend that each influence factor changes over time, and establish influence factor prediction library；
S08, for each history parameters library, calculate its corresponding prediction error, and then miss according to the prediction in N number of period
The changing rule of difference, prediction obtain prediction error and change with time trend, establish prediction error time Sequence Trend library；
S09, it establishes by above-mentioned function data library, Prediction Parameters library, influence factor prediction library and prediction error time sequence
The prediction model that row trend library is formed.
Further, the influence factor include used car supply factor, new car price factor, new model market manifestation because
Element, propensity to consume factor, vehicle depreciation factor, configuration value depreciation factor, city influence factor, type of transaction influence factor and
Individual Effects of Factors factor.
Further, in the step S05, described each vehicle that calculates respectively is in the period interior prediction function and each influence letter
In the step of number corresponding function parameter, the canonical so that being made of anticipation function and each influence function is chosen especially by calculating
Change the parameter value of cost function minimum as corresponding function parameter.
Further, in the step S07, each influence factor is obtained using time series method or causal forecasting model prediction
Change with time trend.
Further, in the step S04, the suitable anticipation function and each chosen from preset alternative function library
In the step of influence function of influence factor, the selection principle that uses for：
Selection of Function forms the combination of function of different anticipation functions and each influence function from preset alternative function library
Afterwards, the price and multiple influence factors that substitution parsing obtains are calculated, the group of functions for finally selecting overall degree of fitting best
It closes.
Further, the historical trading data includes city, vehicle, model, year money, mileage, days of registering the license, color, also
Including type of transaction, transfer number, vehicle condition and/or maintenance situation.
Another technical solution is used by the present invention solves its technical problem：
Automobile residual value Forecasting Methodology based on big data using above-mentioned automobile residual value prediction model, includes the following steps：
S11, the vehicle information and predicted time for obtaining object vehicle input by user；
S12, according to vehicle information input by user, with reference to preset model data library, matching obtains corresponding vehicle and matches
Confidence ceases；
S13, library and prediction error time are predicted from the function data library of prediction model, Prediction Parameters library, influence factor
After corresponding function and parameter are chosen in Sequence Trend library, vehicle information and predicted time are substituted into prediction model and counted
It calculates, obtains prediction residual value of the object vehicle in predicted time；
Wherein, the vehicle information include model, the time of registering the license, mileage, vehicle condition, loco, color, configuration and/or
Physical parameter.
Further, it is further comprising the steps of：
S14, the prediction residual value for obtaining object vehicle different periods in current time to predicted time is calculated, and carries out figure
Shapeization is shown.
The beneficial effects of the invention are as follows：The prediction model that the present invention establishes, including function data library, prediction error time sequence
Row trend library, Prediction Parameters library and influence factor prediction library, have not only fully considered that vehicle factor itself and market factor etc. are each
Influence of the kind influence factor to automobile residual value, while have also contemplated shadow of itself change conditions of these influence factors to automobile residual value
It rings, automobile residual value prediction can be comprehensively carried out with science, improve forecasting accuracy, precision of prediction is high.
Description of the drawings
The automobile residual value prediction model based on big data that Fig. 1 is the present invention establishes process flow diagram flow chart；
The automobile residual value prediction model that Fig. 2 is the present invention establishes principle and application schematic diagram；
Fig. 3 is the flow chart of the automobile residual value Forecasting Methodology based on big data of the present invention.
Specific embodiment
Embodiment one
With reference to Fig. 1, the present invention provides a kind of automobile residual value prediction model based on big data, the prediction model passes through
Following steps, which are established, to be obtained：
S01, a large amount of historical trading datas for obtaining automobile in nearest a period of time；
S02, historical trading data is carried out to intersect data cleaning treatment, rejecting does not have representative historical trading number
According to；
In the present embodiment, it is the logic intersected by various dimensions to judge that every is remembered one by one intersect data cleaning treatment
Whether record is representative, has representative transaction data so as to delete not.The logic of cleaning is exactly whether to judge some vehicle source
It peels off, does not enter modeling if peeling off.And this " group " defines from different dimensions.Such as single dimension occurs 1,000,000
Mileage or 20,000,000 price, this extreme exception record will be cleaned out first.But most of outliers are not so apparent,
Often also need to whether it is abnormal with other dimensions interaction confirmation.If it is 90% singly to see certain vehicle source value preserving rate, it is impossible to is confirmed as
Abnormal point, but divide group according to vehicle age, the Che Yuanwei 3 years is old, 90% value preserving rate inside group of corresponding vehicle age just by
It is considered do not have representative record, is rejected.Here it only lists according to vehicle age and the progress of value preserving rate relationship
The dimension of cleaning.Other relevant dimensions for being related to cleaning also have mileage, body style, vehicle, region etc..
S03, the historical trading data after cleaning is divided into the training data in N number of period, with reference to preset vehicle number
According to library, the training data in each period is resolved into price and the multiple and relevant influence factor of used car price；N is
Natural number；
The process parsed with reference to preset model data library：It is found in model data library by brand matching list first
Corresponding standard brand name.Secondly the vehicle in newly-increased vehicle source and model name can with each vehicle affiliated under the brand and model by
One carries out fuzzy matching, finds the several former candidate model best with vehicle source name matching degree.Then being consistent in conjunction with configuration
The synthesis such as match condition of situation and selling time and year money is given a mark, and score highest and is considered matching higher than threshold value person
Success.It transfers the vehicle of the standard model matched and configuration information fills into newly-increased vehicle source record and completes parsing.Therefore it parses
The information other than price arrived is influence factor signified in the present embodiment.Influence factor be actually and vehicle
Brand, vehicle, configuration, city, time, type of transaction, vehicle age, mileage, vehicle condition, color etc. and the relevant characteristic of price, each
Characteristic may all impact the price of used car different in terms of, these characteristics are summarized as used car supply by the present invention
Factor, new car price factor, new model market manifestation factor, propensity to consume factor, vehicle depreciation factor, configuration value depreciation because
Element, city influence factor, type of transaction influence factor and individual Effects of Factors factor.
S04, the analysis result according to all historical trading datas, for each vehicle, from preset alternative function library
Choose suitable anticipation function and the influence function of each influence factor, component function database；
S05, the analysis result based on the training data in each period are calculated respectively using optimization algorithm so that predicting
The function parameter combination of the cost function of function and each influence function minimum, so as to establish history parameters corresponding with the period
Library；
The Computing Principle of step S05 be first set with anticipation function and the relevant cost function of each influence function, such as
Set residual sum of squares (RSS) of the cost function as anticipation function and each influence function.Then for the prediction letter chosen in step S04
The combination of function of number and each influence function, based on the analysis result of the training data in each period, using optimization algorithm meter
It calculates and selects one group of function parameter combination so that cost function of each vehicle within the period is minimum.Here function parameter
Combination includes the function parameter of anticipation function and each influence function, finally, by the corresponding function parameter of each period of acquisition
History parameters library corresponding with the period is established in combination, i.e., for each vehicle, is respectively provided with history corresponding with multiple periods
Parameter library, so as to which subsequent step can obtain function parameter at any time according to corresponding history parameters library of multiple periods to predict
Between variation tendency.
S06, for each vehicle, according to its different history parameters library within N number of period, prediction obtains different parameters
Change with time trend, establishes Prediction Parameters library；Specifically, anticipation function, influence function for each vehicle is each
Parameter sees that each parameter, which changes over time, has N number of value along time shaft, therefore, can be according to the situation of this N number of value, in advance
It surveys and obtains parameter and change with time trend, and then after obtaining the Trend Forecast of all parameters, Prediction Parameters can be established
Library.Specifically, can the parameter value situation in following 1-5 be predicted according to the situation of N number of value.If for example, the three of GPS
It is 0.35,17 years is 0.32 that yearly depreciation, which was 0.4,16 years in 15 years, then is not difficult to deduce that future will continue to drop.If use logarithm
Curve matching can then obtain prediction model as allowance for depreciation=- 0.073*ln (prediction time -2014)+0.4001.Then thus may be used
To deduce in Prediction Parameters library, tri- yearly depreciations of GPS correspond to 0.30/0.28/0.27 respectively in 2018-2020.
During practical progress parameter prediction, a variety of songs such as linear, index, logarithm, hyperbolic, multinomial, trigonometric function
Line can change with time situation for fitting parameter, for different parameters, different curves can be selected to be intended
It closes, Prediction Parameters library is established according to the parameter prediction result that the best function of fitting effect and parameter calculate.
S07, prediction obtain the trend that each influence factor changes over time, and establish influence factor prediction library；
S08, for each history parameters library, calculate its corresponding prediction error, and then miss according to the prediction in N number of period
The changing rule of difference, prediction obtain prediction error and change with time trend, establish prediction error time Sequence Trend library；
Such as prediction error time sequence shows that A vehicles can be underestimated 2% in each 12 monthly average, then the result will be by
For adjusting the prediction (such as the predicted value in December is multiplied by 1.02) in the vehicle following all December.Certainly here
What example was enumerated is simple rule, rule or even trend that in practice can be more complicated, but principle is identical.By this step,
The prediction error that prediction obtains this model changes with time, and so as to be modified to this prediction model, can improve this prediction
The precision of prediction of model.
S09, it establishes by above-mentioned function data library, Prediction Parameters library, influence factor prediction library and prediction error time sequence
The prediction model that row trend library is formed.
Fig. 2 is that this prediction model establishes principle and application schematic diagram, in figure, F (T, X1, X2, X3 ..., Xn |
β1, β2, β3...) and represent prediction model anticipation function, wherein, X1, X2, X3 ... ..., Xn represent each influence factor respectively
Influence function, transferred from function data library, and impacted factor prediction library influence, T represent prediction error, from prediction miss
It is obtained in poor time series trend library, β1, β2, β3The function parameter of influence function obtained in Prediction Parameters library is represented respectively.It is logical
Cross after establishing this prediction model, can with science, comprehensively carry out automobile residual value prediction, accuracy is high.
Be further used as preferred embodiment, the influence factor include used car supply factor, new car price factor,
New model market manifestation factor, vehicle depreciation factor, configuration value depreciation factor, city influence factor, is handed over propensity to consume factor
Easy patterns affect factor and individual Effects of Factors factor.
Used car supply factor refers to influence of the used car supply to automobile residual value, because of the used car supply of certain a automobile
The difference of amount can have automobile residual value Different Effects, therefore, when carrying out the prediction of automobile residual value, need to fully consider that used car is supplied
The influence of amount factor.Similarly, new car price factor, new model market manifestation factor, propensity to consume factor reflect difference respectively
Influence of the factor to used car residual value so when carrying out residual value prediction, needs to consider comprehensively.
Corresponding, Prediction Parameters library includes vehicle depreciation parameter library, configuration value depreciation parameter library, city affecting parameters
Library, type of transaction affecting parameters library, individual Effects of Factors parameter library and used car supply parameter library, new car price parameter library,
New model market manifestation parameter library, propensity to consume parameter library.
Specifically, in the present embodiment, model data library is configured as：The characteristic information of each vehicle models is included, such as
The affiliated vehicle of the model and brand, new car period on sale, year money, manufacturer's recommendations valency, important configuration information and physics ginseng
The information such as number length, width and height discharge capacity.
Each parameter library in Prediction Parameters library is described in detail as follows：
First, vehicle standard depreciation library describes the depreciation curve that each vehicle increases according to vehicle age.For different cities this
A depreciation curve can be different.Then during prediction model is established, for different cities, the parameter in vehicle standard depreciation library
It can be different.
2nd, configuration value depreciation library：Describe each distinctive depreciation curve of emphasis configuration information of vehicle, this depreciation
Curve can be identical with vehicle depreciation curve, can also be different.
3rd, city affecting parameters library：Record each city the relatively national average overall offset situation of used car price,
Which and these vehicle individual offset situations the special vehicle not being consistent with overall offset in the city be.
4th, type of transaction affecting parameters library：Record different type of transaction price differential for caused by price.This type of transaction
Price differential can change with urban changes.
5th, individual Effects of Factors parameter library：The parameter and function of all kinds of individual Effects of Factors are included, these functions can incite somebody to action
The situation of change conversion of the individual factor such as mileage, time of registering the license, vehicle condition, color is the value to used car price.Different cities
This set of affecting parameters of city's different automobile types can be different.
Specific in step S07, because each influence factor all has an impact the residual value of automobile, therefore, the present embodiment removes
It being fully considered when establishing prediction model except each influence factor, it is also contemplated that each influence factor changes with time trend,
Prediction obtains the Trend Forecast of each influence factor, and used car supply anticipation trend as shown in Figure 2, the propensity to consume are pre-
Survey trend and new car price expectation trend and new car market manifestation anticipation trend etc., so as to which these anticipation trends be established
Influence factor predicts library, so as to be modified to prediction model so that prediction result is more accurate.
Wherein, used car supply anticipation trend is used for：The variation tendency of the following used car supply of prediction；
New car price expectation trend is used for：The variation tendency of the following new car price of prediction；
Propensity to consume anticipation trend is used for：The different propensity to consume variation tendencies on used-vehicle market that segment market of prediction；
It segments market and refers to the segmenting market for vehicle, vehicle similar in same class price function segments market, such as tightly referred to as one
Gather type SUV, intermediate-size car, compact car etc. is exactly different segments market.
New model market manifestation factor is used for：The market of this following vehicle of the prediction such as product feature according to new listing vehicle
Performance trend.
It is assumed that influence of the known used car supply to residual value is F (used car supply variable quantity * β), and Prediction Parameters library
The parameter beta of offer shows that used car supply often doubles, and automobile residual value can reduce 3%.But predict the year two thousand twenty supply pair
The influence of residual value, it will also be appreciated that the change conditions of following used car supply.Therefore it needs to know that the year two thousand twenty used car supply can increase
It adds few.It is exactly the result provided by used car supply anticipation trend that this, which increases how many,.Such as the trend provides used car supply
Amount can increase by 50% in the year two thousand twenty than now, then understand that, in the year two thousand twenty, automobile residual value can reduce 1.5% because supply increases
(=3%*50%).
Generally speaking, this prediction model has fully considered the various influence factors such as vehicle factor itself and the market factor to vapour
The influence of vehicle residual value, while the influence of these influence factors itself change conditions to automobile residual value is had also contemplated, it can be complete with science
Automobile residual value prediction is carried out to face, improves forecasting accuracy, precision of prediction is high.
It is further used as preferred embodiment, in the step S05, described each vehicle that calculates respectively is within the period
In the step of anticipation function and each influence function corresponding function parameter, chosen especially by calculating so that by anticipation function and each
The parameter value of the regularization cost function minimum of influence function composition is as corresponding function parameter.
It is further used as preferred embodiment, it is pre- using time series method or causal forecasting model in the step S07
It surveys and obtains each influence factor and change with time trend.
Time series method：Some are difficult to establish causal influence factor with present case, generally passes through classics
Time Series Method is disassembled season, and trend and irregular item, find its trend or rule predicts future.For carrying weight
The economic class variable such as GDP of prediction directly quotes result and changes with time trend as influence factor.
Causal forecasting model：The factor that makes some difference and the current even same class of history or in addition a kind of influence factor have because
Fruit relationship, this kind of influence factor can be predicted by establishing the model of itself and current influence factor quantitative relationship.Such as
The sale situation of used car supply and in history new car is closely coupled, and it may be said that the sales volume of new car is the drive of used car supply
Dynamic variable.So the causal forecasting model by establishing current used car supply and the passing sales volume of new car in history, obtains used car
Supply anticipation trend.After this causal forecasting model establishes, following used car can be supplied by existing new car sales volume
Amount is predicted.Such as be fitted to obtain following used car supply anticipation trend by historical data be：3 years T periods used car is supplied
Amount=(T-3) period new car sales volume * 0.17.
It is further used as preferred embodiment, in the step S04, described chosen from preset alternative function library is closed
In the step of suitable anticipation function and the influence function of each influence factor, the selection principle that uses for：
Selection of Function forms the combination of function of different anticipation functions and each influence function from preset alternative function library
Afterwards, the price and multiple influence factors that substitution parsing obtains are calculated, the group of functions for finally selecting overall degree of fitting best
It closes.
Further, it is noted that anticipation function that the present embodiment to each vehicle, selects overall degree of fitting best and multiple
Influence function, so as to which in the prediction model of structure, each vehicle is respectively provided with the prediction letter that acquisition is selected according to best fitted principle
Several and corresponding influence function is predicted with reference to the history parameters library that acquisition is calculated according to the training data of each period and is obtained
The Prediction Parameters library obtained, the present invention can obtain each vehicle prediction result the most accurate.
In the present embodiment, the alternative function library is included with minor function：Linear function, polynomial function, exponential function
Race, piecewise function, trigonometric function, hyperbolic functions and/or indicative function
It is further used as preferred embodiment, the historical trading data includes city, vehicle, model, year money, inner
Journey, days of registering the license, color further include type of transaction, transfer number, vehicle condition and/or maintenance situation.
Embodiment two
With reference to Fig. 3, the automobile residual value Forecasting Methodology based on big data predicts mould using the automobile residual value of above-described embodiment one
Type includes the following steps：
S11, the vehicle information and predicted time for obtaining object vehicle input by user；In this step, it can pass through
APP, website or api interface obtain data input by user；
S12, according to vehicle information input by user, with reference to preset model data library, matching obtains corresponding vehicle and matches
Confidence ceases；
S13, library and prediction error time are predicted from the function data library of prediction model, Prediction Parameters library, influence factor
After corresponding function and parameter are chosen in Sequence Trend library, vehicle information and predicted time are substituted into prediction model and counted
It calculates, obtains prediction residual value of the object vehicle in predicted time；
Wherein, the vehicle information include model, the time of registering the license, mileage, vehicle condition, loco, color, configuration and/or
Physical parameter.
The above-mentioned prediction model of this method obtains automobile residual value to predict, science, comprehensive, prediction accuracy height, and stablizes
Property it is high, can relatively accurately predict and obtain automobile residual value.
In practical calculating, after step S12 matchings obtain corresponding vehicle configuration information, vehicle information and vehicle configuration are believed
After breath is converted to numerical variable, then substitutes into anticipation function and calculated.Available data may be used in specifying information conversion process
Common practice in processing identifies some important informations for example, by using identification code, and time of registering the license etc. is identified using total duration.
Preferred embodiment is further used as, it is further comprising the steps of：
S14, the prediction residual value for obtaining object vehicle different periods in current time to predicted time is calculated, and carries out figure
Shapeization is shown.
The method can calculate the bus predicted residual value obtained in a period of time, and pass through the graphic software platforms shape such as curve
The dynamic change of formula display automobile residual value, can intuitively, explicitly indicate residual value anticipation trend.
Embodiment three
The present embodiment is the detailed example of embodiment two, specifically includes step：
Step 1, the vehicle information for obtaining object vehicle input by user and predicted time are as follows：
Model：2017 sections of 200 4MATIC of C L movement versions of BeiJing BenChi's benz C grades；
Place：Beijing；
Color：White；
It registers the license the time：In October, 2017；
Predict used car on-sale date：The year two thousand twenty October.
Object vehicle described in the present embodiment refers to the used car of residual value prediction to be carried out.
Step 2, according to vehicle information input by user, with reference to preset model data library, matching obtain it is corresponding including
Vehicle configuration information including Standard of vehicle model：
Discharge capacity：2.0；
Horsepower：184；
Engine type：Turbocharging；
Gearbox-type：Automatically；
Type of drive：Full-time four-wheel drive；
Discharge capacity：State five；
MSRP:39.98 ten thousand.MSRP represents manufacturer's recommendations retail price.
Step 3, setting below in relation to service condition influence factor：
Service condition：Non- operation；
Mileage：4.5 ten thousand kilometers；
Vehicle condition：It is outstanding；
Type of transaction：Manufacturer auctions off to itself second-hand car trader；
Transfer ownership number：0.
Step 4 calculates influence of each influence factor to automobile residual value.By from the function data library of prediction model, prediction
It, will after choosing corresponding function and parameter in parameter library, influence factor prediction library and prediction error time Sequence Trend library
Vehicle information and predicted time substitute into prediction model and are calculated, and obtain prediction residual value of the object vehicle in predicted time.
A, zequin：：MSRP 39.98 ten thousand；
B, anticipation function is inquired：Pass through the corresponding anticipation function of function data library lookup of prediction model, index key
For：Benz C+ Guangzhou, obtaining anticipation function is：
(MSRP+ time effects * configuration influence+supply influence+year money influence+propensity to consume influences+mileage of registering the license influence
+ type of transaction influences+other individual influences) * Color influences * cities influence * exchange hours influence+model error adjustment item
It is noted that the anticipation function of the present invention not only only has the combination that above-mentioned adduction multiplies, it is possibility to have other combinations
Mode, such as exponential form, logarithmic form etc., the present embodiment does not enumerate.
C, obtaining time effects price of registering the license according to time calculating of registering the license is：- 9 ten thousand, specially：
1) vehicle standard depreciation library, is looked into, index key is：Run quickly C, and 2017 sections, Beijing, 36 months, the year two thousand twenty obtained
＆ month vehicle ages in year influence coefficient and influence function G (), and it is then the function parameter of influence function that vehicle age, which influences coefficient, here.
2) influence calculating of the time for used car price of registering the license is carried out：
G (36 months * month vehicle ages influenced coefficient, and 3 years * vehicle ages influenced coefficient)=- 9 ten thousand
D, according to vehicle configuration, " run quickly C, Beijing, 36 months, the year two thousand twenty " is searched in configuration value depreciation library, is obtained
Influence value is " 110% ".
F, the influence price for calculating used car supply is -1.05 ten thousand.Calculating process is as follows：
First, it is matched in influence factor predicts library, index key is：Run quickly C, Beijing, and the year two thousand twenty is supplied
Measure increasing degree：70%；
2nd, it is matched in Prediction Parameters library, index key is：Run quickly C, Beijing, the year two thousand twenty, obtains used car supply
Measuring affecting parameters is：- 1.5 ten thousand/supplys often increase by 100%；
3rd, the influence of used car supply is calculated：70%*-1.5 ten thousand=- 1.05 ten thousand.
G, calculating acquisition influence price according to 3 years used car moneys is：- 0.3 ten thousand, process is similar to above-mentioned steps c, also in
Vehicle standard depreciation is matched in library, and index key is：Run quickly C, Beijing, the year two thousand twenty.
H, it is equally matched in vehicle standard depreciation library, index key is：Run quickly C, Beijing, the year two thousand twenty, matching
The vehicle is obtained as latest generation, at a discount+1.2 ten thousand.
I, the influence price that calculating obtains the propensity to consume is：+ 0.8 ten thousand.In influence factor predicts library and Prediction Parameters library
It is matched, calculation is similar with step f.
J, according to mileage, individual Effects of Factors parameter library using index key " benz C, Beijing, the year two thousand twenty, 4.5 ten thousand
It is -1.3 ten thousand at a discount that kilometer " matching, which obtains,.
K, according to type of transaction, " benz C, Beijing, the year two thousand twenty, manufacturer's bat in type of transaction affecting parameters library, are searched
Sell ", it obtains being -2.1 ten thousand at a discount.
L, it to other individual factors, is searched in individual Effects of Factors parameter library and obtains corresponding being at a discount 0.7 ten thousand.Specifically
Individual factors include vehicle condition, transfer ownership number, and whether quality guarantee expire, repair etc..
M, according to vehicle color, in individual Effects of Factors parameter library using index key " benz C, Beijing, 2020
Year, white " acquisition influence value is matched as " 98% ".
According to city, " benz C, Beijing, the year two thousand twenty " in the affecting parameters library of city, is searched, obtaining influence value is
" 97% ".
According to the time, " benz C, Beijing, the year two thousand twenty " in time influencing parameters library, is searched, obtaining influence value is
" 96% ".
Predict error transfer factor case：In prediction error time Sequence Trend library, " benz C, Beijing, the year two thousand twenty " is searched, is obtained
It is -0.5 ten thousand to price is influenced.
N, by the influence result of above-mentioned influence factor together calculate obtain the used car residual value be：
(39.98-9*110%-1.05-0.3+1.2+0.8-1.3-2.1+0.7) * 98%*97%*96%-0.5=
25.08 ten thousand.
Therefore, this Forecasting Methodology has fully considered that the various influence factors such as vehicle factor itself and the market factor are residual to automobile
The influence of value, while the influence of these influence factors itself change conditions to automobile residual value is had also contemplated, it can be with science comprehensively
Automobile residual value prediction is carried out, improves forecasting accuracy, precision of prediction is high.
It is that the preferable of the present invention is implemented to be illustrated, but the invention is not limited to the implementation above
Example, those skilled in the art can also make various equivalent variations under the premise of without prejudice to spirit of the invention or replace
It changes, these equivalent modifications or replacement are all contained in the application claim limited range.
Claims (8)
1. the automobile residual value prediction model based on big data, which is characterized in that the prediction model is obtained by following steps foundation
：
S01, a large amount of historical trading datas for obtaining automobile in nearest a period of time；
S02, historical trading data is carried out to intersect data cleaning treatment, rejecting does not have representative historical trading data；
S03, the historical trading data after cleaning is divided into the training data in N number of period, with reference to preset model data library,
Training data in each period is resolved into price and the multiple and relevant influence factor of used car price；N is nature
Number；
S04, the analysis result according to all historical trading datas for each vehicle, are chosen from preset alternative function library
The influence function of suitable anticipation function and each influence factor, component function database；
S05, the analysis result based on the training data in each period are calculated respectively using optimization algorithm so that anticipation function
It is combined with the function parameter of the cost function minimum of each influence function, so as to establish history parameters library corresponding with the period；
S06, for each vehicle, according to its different history parameters library within N number of period, prediction obtains different parameters at any time
Between variation tendency, establish Prediction Parameters library；
S07, prediction obtain the trend that each influence factor changes over time, and establish influence factor prediction library；
S08, for each history parameters library, calculate its corresponding prediction error, and then according to the prediction error in N number of period
Changing rule, prediction obtain prediction error and change with time trend, establish prediction error time Sequence Trend library；
S09, it establishes and is become by above-mentioned function data library, Prediction Parameters library, influence factor prediction library and prediction error time sequence
The prediction model that gesture library is formed.
2. the automobile residual value prediction model according to claim 1 based on big data, which is characterized in that the influence factor
Including used car supply factor, new car price factor, new model market manifestation factor, propensity to consume factor, vehicle depreciation factor,
Configuration value depreciation factor, city influence factor, type of transaction influence factor and individual Effects of Factors factor.
3. the automobile residual value prediction model according to claim 2 based on big data, which is characterized in that the step S05
In, described each vehicle that calculates respectively in the step of period interior prediction function and each influence function corresponding function parameter,
It is chosen especially by calculating so that being made by the parameter value of regularization cost function minimum that anticipation function and each influence function form
For corresponding function parameter.
4. the automobile residual value prediction model according to claim 1 based on big data, which is characterized in that the step S07
In, each influence factor is obtained using time series method or causal forecasting model prediction and is changed with time trend.
5. the automobile residual value prediction model according to claim 1 based on big data, which is characterized in that the step S04
In, it is described suitable anticipation function and each influence factor are chosen from preset alternative function library influence function the step of
In, the selection principle that uses for：
After Selection of Function forms the combination of function of different anticipation functions and each influence function from preset alternative function library, generation
Enter to parse obtained price and multiple influence factors are calculated, the combination of function for finally selecting overall degree of fitting best.
6. the automobile residual value prediction model according to claim 1 based on big data, which is characterized in that the historical trading
Data include city, vehicle, model, year money, mileage, days of registering the license, color, further include type of transaction, transfer number, vehicle condition
And/or maintenance situation.
7. the automobile residual value Forecasting Methodology based on big data predicts mould using claim 1-6 any one of them automobiles residual value
Type, which is characterized in that include the following steps：
S11, the vehicle information and predicted time for obtaining object vehicle input by user；
S12, according to vehicle information input by user, with reference to preset model data library, matching obtains corresponding vehicle configuration letter
Breath；
S13, library and prediction error time sequence are predicted from the function data library of prediction model, Prediction Parameters library, influence factor
After corresponding function and parameter are chosen in trend library, vehicle information and predicted time are substituted into prediction model and calculated, is obtained
Obtain prediction residual value of the object vehicle in predicted time；
Wherein, the vehicle information includes model, time of registering the license, mileage, vehicle condition, loco, color, configuration and/or physics
Parameter.
8. the automobile residual value Forecasting Methodology according to claim 7 based on big data, which is characterized in that further include following step
Suddenly：
S14, the prediction residual value for obtaining object vehicle different periods in current time to predicted time is calculated, and be patterned
Display.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711479877.3A CN108154275A (en) | 2017-12-29 | 2017-12-29 | Automobile residual value prediction model and Forecasting Methodology based on big data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711479877.3A CN108154275A (en) | 2017-12-29 | 2017-12-29 | Automobile residual value prediction model and Forecasting Methodology based on big data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108154275A true CN108154275A (en) | 2018-06-12 |
Family
ID=62460117
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711479877.3A Pending CN108154275A (en) | 2017-12-29 | 2017-12-29 | Automobile residual value prediction model and Forecasting Methodology based on big data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108154275A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109345136A (en) * | 2018-10-22 | 2019-02-15 | 中山大学 | Vehicle survival curve model optimization method based on China's motor vehicle resignation system |
| CN110348880A (en) * | 2019-06-18 | 2019-10-18 | 深圳市元征科技股份有限公司 | Assess the method and device of the rental price of vehicle |
| CN110555024A (en) * | 2019-08-13 | 2019-12-10 | 广东数鼎科技有限公司 | Accurate automobile model matching system based on artificial intelligence algorithm |
| CN110599230A (en) * | 2018-06-13 | 2019-12-20 | 优估（上海）信息科技有限公司 | Method for constructing pricing model of second-hand vehicle, pricing method and device |
| CN110648154A (en) * | 2018-06-26 | 2020-01-03 | 优信拍（北京）信息科技有限公司 | Estimation method and system for second-hand car value-preserving rate and service server |
| CN110706039A (en) * | 2019-10-16 | 2020-01-17 | 爱驰汽车有限公司 | Electric vehicle residual value rate evaluation system, method, equipment and medium |
| CN112070535A (en) * | 2020-09-03 | 2020-12-11 | 常州微亿智造科技有限公司 | Electric vehicle price prediction method and device |
| CN112101998A (en) * | 2020-09-15 | 2020-12-18 | 北京嘀嘀无限科技发展有限公司 | Residual value determining method, model obtaining method, device, equipment and storage medium |
| CN112579900A (en) * | 2020-12-22 | 2021-03-30 | 优必爱信息技术（北京）有限公司 | Method, system and equipment for recommending second-hand vehicle replacement information |
| CN113329021A (en) * | 2021-05-31 | 2021-08-31 | 湖北工业大学 | Automobile depreciation model parameter privacy protection system and method based on industrial Internet |
| CN114331573A (en) * | 2022-03-15 | 2022-04-12 | 蜗牛货车网(山东)电子商务有限公司 | Vehicle residual value evaluation method based on big data and trading platform |
| CN114997756A (en) * | 2022-08-03 | 2022-09-02 | 成都智暄科技有限责任公司 | Vehicle residual value evaluation system and method based on cockpit data |
-
2017
- 2017-12-29 CN CN201711479877.3A patent/CN108154275A/en active Pending
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110599230A (en) * | 2018-06-13 | 2019-12-20 | 优估（上海）信息科技有限公司 | Method for constructing pricing model of second-hand vehicle, pricing method and device |
| CN110599230B (en) * | 2018-06-13 | 2022-07-22 | 优轩(北京)信息科技有限公司 | Second-hand car pricing model construction method, pricing method and device |
| CN110648154A (en) * | 2018-06-26 | 2020-01-03 | 优信拍（北京）信息科技有限公司 | Estimation method and system for second-hand car value-preserving rate and service server |
| WO2020082683A1 (en) * | 2018-10-22 | 2020-04-30 | 中山大学 | Chinese motor vehicle registration system-based method for optimizing vehicle survival curve model |
| CN109345136A (en) * | 2018-10-22 | 2019-02-15 | 中山大学 | Vehicle survival curve model optimization method based on China's motor vehicle resignation system |
| CN110348880A (en) * | 2019-06-18 | 2019-10-18 | 深圳市元征科技股份有限公司 | Assess the method and device of the rental price of vehicle |
| CN110555024B (en) * | 2019-08-13 | 2022-03-29 | 广东数鼎科技有限公司 | Accurate automobile model matching system based on artificial intelligence algorithm |
| CN110555024A (en) * | 2019-08-13 | 2019-12-10 | 广东数鼎科技有限公司 | Accurate automobile model matching system based on artificial intelligence algorithm |
| CN110706039A (en) * | 2019-10-16 | 2020-01-17 | 爱驰汽车有限公司 | Electric vehicle residual value rate evaluation system, method, equipment and medium |
| CN112070535A (en) * | 2020-09-03 | 2020-12-11 | 常州微亿智造科技有限公司 | Electric vehicle price prediction method and device |
| CN112101998A (en) * | 2020-09-15 | 2020-12-18 | 北京嘀嘀无限科技发展有限公司 | Residual value determining method, model obtaining method, device, equipment and storage medium |
| CN112579900A (en) * | 2020-12-22 | 2021-03-30 | 优必爱信息技术（北京）有限公司 | Method, system and equipment for recommending second-hand vehicle replacement information |
| CN112579900B (en) * | 2020-12-22 | 2024-03-12 | 优必爱信息技术（北京）有限公司 | Second-hand vehicle replacement information recommendation method, system and equipment |
| CN113329021A (en) * | 2021-05-31 | 2021-08-31 | 湖北工业大学 | Automobile depreciation model parameter privacy protection system and method based on industrial Internet |
| CN113329021B (en) * | 2021-05-31 | 2022-04-29 | 湖北工业大学 | Automobile depreciation model parameter privacy protection system and method based on industrial Internet |
| CN114331573A (en) * | 2022-03-15 | 2022-04-12 | 蜗牛货车网(山东)电子商务有限公司 | Vehicle residual value evaluation method based on big data and trading platform |
| CN114331573B (en) * | 2022-03-15 | 2022-06-21 | 蜗牛货车网(山东)电子商务有限公司 | Vehicle residual value evaluation method based on big data and transaction platform |
| CN114997756A (en) * | 2022-08-03 | 2022-09-02 | 成都智暄科技有限责任公司 | Vehicle residual value evaluation system and method based on cockpit data |
| CN114997756B (en) * | 2022-08-03 | 2022-12-20 | 成都智暄科技有限责任公司 | Vehicle residual value evaluation system and method based on cockpit data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108154275A (en) | Automobile residual value prediction model and Forecasting Methodology based on big data | |
| CN108280674A (en) | A kind of method and system carrying out used car valuation using big data | |
| JP5963861B2 (en) | System and method for analysis and presentation of used vehicle pricing data | |
| JP5566422B2 (en) | High risk procurement analysis and scoring system | |
| US11410206B2 (en) | Systems and methods for transformation of raw data to actionable data | |
| US20170300991A1 (en) | Used-vehicle algorithmic pricing engine method and system | |
| US20070250401A1 (en) | Searching method and apparatus | |
| US20080027882A1 (en) | Price assessment method for used equipment | |
| JP6727356B2 (en) | Patent evaluation system | |
| US20240211977A1 (en) | System and method for correlating and enhancing data obtained from distributed sources in a network of distributed computer systems | |
| CN111768243A (en) | Sales prediction method, prediction model construction method, device, equipment and medium | |
| CA2881812A1 (en) | System, method and computer program for forecasting residual values of a durable good over time | |
| CN115310256A (en) | Vehicle estimation method and device | |
| CN115147144A (en) | Data processing method and electronic equipment | |
| CN114943582A (en) | Information recommendation method and system and recommendation server | |
| CN114707771A (en) | Big data based medicine supply chain management method and platform | |
| JP4921572B2 (en) | Bond characteristic calculation system and spread change rate calculation system | |
| CN116304374B (en) | Customer matching method and system based on package data | |
| CN117172825A (en) | Electronic commerce data analysis method and device based on big data and computing equipment | |
| CN107093103B (en) | Brand value evaluation method and system based on big data statistical analysis | |
| JP2017146632A (en) | Replacement timing determining method, replacement timing determining system, and program | |
| Akintoye | Construction tender price index: modelling and forecasting trends | |
| CN110968622B (en) | Accounting report customization method, platform and terminal | |
| Milošević et al. | Determination of residual value of construction machinery based on machine age | |
| US20130006704A1 (en) | Method and System For Creating Customer Profiles |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180612 |
|
| RJ01 | Rejection of invention patent application after publication |