KR20230034608A - Method, device and system for product sales forecasting and product recommendation information generation using artificial intelligence based on sales data - Google Patents

Abstract

A method for predicting product sales and generating product recommendation information using artificial intelligence based on sales data comprises the steps of: obtaining daily sales quantity information of a first product up to a first time point; generating a first prediction model based on artificial intelligence by using the obtained daily sales quantity information as learning data in order to predict the sales quantity of the first product at a second time point, wherein the second time point is a time point after the first time point; predicting the sales quantity of the first product at the second time point by using the generated first prediction model based on the daily sales quantity information; and selecting one or more recommended products by using the previously generated first prediction model. A device according to one embodiment may be combined with hardware and controlled by a computer program stored in a medium to execute any one of the above-described methods.

Description

Product sales forecasting and product recommendation information generation method, device and system using artificial intelligence based on sales data

The following embodiments relate to a method, apparatus, and system for predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data.

With the development of the Internet and mobile communication, the use of online shopping malls for both B2B and B2C transactions is increasing.

As the product recommendation system is advanced in online e-commerce, it is possible to actively cross-sell or up-sell by recommending products suitable for the customer's taste, and as a result, the popular trend of Long-Tail The development of social network services has brought changes to online purchasing attitudes, and the authority of experts and the influence of advertisements are gradually weakening, and the influence of recommendation systems through personalization is gradually increasing in product purchases.

In addition, proper inventory management is important for sales of products such as clothing. In particular, based on the warehousing period after production or ordering, inventory corresponding to the expected sales within the period must be held to respond to the buyer's request in real time. If you fail to meet customer orders due to insufficient inventory, you not only lose sales opportunities, but also risk customer churn. In order to solve this problem, it is required to maintain an appropriate inventory level, and for this purpose, it is necessary to accurately predict future product sales volume.

In general, it takes a certain period of time from a purchase order to actual goods being delivered. Therefore, the above problem occurs if an additional quantity is not ordered in advance based on the expected sales volume at a future point in time. In particular, in the case of externally produced products, there is a risk that the required quantity cannot be secured fundamentally unless the stock is replenished in advance.

Although it is necessary to forecast product sales volume to maintain adequate inventory as described above, the reason why the forecast was not properly made is as follows. First, product sales volume is affected by various variables and is difficult to predict because it is exposed to constant uncertainty. In addition, in the case of an online shopping mall, since the shopping mall operator does not have professional knowledge to predict product sales volume and has to manage a large number of products, there is a problem in that excessive cost and effort are required to predict the sales volume of all products. In addition, there are cases in which the operator's arbitrary judgment intervenes. In this case, there is a problem in that proper inventory management is not performed because decision-making based on the same standard is not made for each product.

Embodiments are to provide a method, apparatus, and system for predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data.

A method of predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data, the method comprising: obtaining daily sales quantity information of a first product up to a first point in time; generating an artificial intelligence-based first prediction model by using the obtained daily sales quantity information as learning data to predict the sales quantity of the first product at a second time point, wherein the second time point is the first is a point in time after a point in time; predicting the sales quantity of the first product at the second time point using the first prediction model generated based on the daily sales quantity information; and selecting one or more recommended products using a pre-generated first prediction model. Based on sales data including, it provides a method of predicting product sales volume and generating product recommendation information using artificial intelligence.

An apparatus according to an embodiment may be combined with hardware and controlled by a computer program stored in a medium to execute any one of the methods described above.

The embodiments have an effect of efficiently providing a method, apparatus, and system for predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data.

1 is a flowchart illustrating a method of predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data.
2 is a diagram showing an LSTM (Long Short Term Memory) model applicable to a method of predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data according to an embodiment of the present invention by way of example.
3 is an exemplary diagram of a configuration of a device according to an embodiment.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, since various changes can be made to the embodiments, the scope of rights of the patent application is not limited or limited by these embodiments. It should be understood that all changes, equivalents or substitutes to the embodiments are included within the scope of rights.

Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only, and may be modified and implemented in various forms. Therefore, the embodiments are not limited to the specific disclosed form, and the scope of the present specification includes changes, equivalents, or substitutes included in the technical spirit.

Although terms such as first or second may be used to describe various components, such terms should only be construed for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element.

It should be understood that when an element is referred to as being “connected” to another element, it may be directly connected or connected to the other element, but other elements may exist in the middle.

Terms used in the examples are used only for descriptive purposes and should not be construed as limiting. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, terms such as "include" or "have" are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art to which the embodiment belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and unless explicitly defined in the present application, they should not be interpreted in an ideal or excessively formal meaning. don't

In addition, in the description with reference to the accompanying drawings, the same reference numerals are given to the same components regardless of reference numerals, and overlapping descriptions thereof will be omitted. In describing the embodiment, if it is determined that a detailed description of a related known technology may unnecessarily obscure the gist of the embodiment, the detailed description will be omitted.

The embodiments may be implemented in various types of products such as personal computers, laptop computers, tablet computers, smart phones, televisions, smart home appliances, intelligent vehicles, kiosks, and wearable devices.

1 is a flowchart illustrating a method of predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data.

A method for predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data according to an aspect of the present invention for achieving the above object is to obtain daily sales quantity information of a first product up to a first point in time generating an artificial intelligence-based first prediction model by using the obtained daily sales quantity information as learning data to predict the sales quantity of the first product at the second time point (the second time point is predicting the sales quantity of the first product at the second time point using the generated first prediction model based on the first point in time) and the daily sales quantity information; The generating of the first prediction model may include: learning the first prediction model while changing a value of a hyper parameter of the first prediction model using the daily sales quantity information; and determining a hyper parameter value having the best learning performance as a hyper parameter value of the first prediction model as a result of evaluating learning performance using the daily sales quantity information.

The method may further include determining whether to use the daily sales quantity information as learning data by determining whether there is no sales quantity information on a specific date among the daily sales quantity information of the first product.

The product sales volume prediction method includes comparing the sales quantity value for each day of the daily sales quantity information with the median value by using a median filter, and as a result of the comparison, at a third point in time when the value deviated from the median value is greater than the threshold value. The method may further include selecting a daily sales quantity value and replacing the selected daily sales quantity value at the third time point with a value calculated using a median value.

In the step of replacing the value of the daily sales volume at the selected third time point with a value calculated using the median value, the daily sales volume value at the selected third time point is replaced with the average sales volume up to the day before the third time point. Afterwards, a step of adding an amount equal to the difference from the replaced value may be included.

The product sales volume prediction method may further include performing standard scaling on daily sales quantity values having values processed through a median filter.

The generating of the first prediction model may include generating a plurality of structured candidate prediction models by combining at least some of a plurality of previously held AI-based prediction model sequences, the plurality of structured candidate prediction models. inputting and learning a first part of the daily sales quantity information as learning data, evaluating the learning performance of the plurality of structured candidate prediction models using a second part of the daily sales quantity information, and evaluating the evaluation result The method may include selecting a structured candidate prediction model having the best learning performance as the first prediction model.

The generating of the first prediction model may include generating a plurality of structured candidate prediction models and learning the plurality of structured candidate prediction models by inputting a first part of the daily sales quantity information as training data; Evaluating the learning performance of the plurality of structured candidate prediction models by extracting and using a second part of the daily sales quantity information as test data, and evaluating the structured candidate prediction model having the best learning performance as a result of the evaluation. A step of selecting a predictive model may be further included.

The plurality of artificial intelligence-based predictive model sequences have hyperparameter values corresponding to each sequence, and the hyperparameters corresponding to each sequence include a first variable associated with the number of memory cells and a second variable associated with the number of input or output attributes. It may include at least one of variables and a third variable related to the input or output length of time series data.

The generating of the first prediction model may include learning the first prediction model while changing a value of a hyper parameter of the first prediction model using the first part of the daily sales quantity information. and as a result of evaluating learning performance using the second part of the daily sales quantity information, determining a hyperparameter value having the best learning performance as a hyperparameter value of the first prediction model. can

A product sales volume prediction method according to an aspect of the present invention for achieving the above object includes obtaining daily sales quantity information of a first product up to a first time point, predicting the sales quantity of the first product at a second time point generating a first artificial intelligence-based prediction model using the obtained daily sales quantity information as learning data (the second time point is a time point after the first time point) and the daily sales quantity information predicting the sales quantity of the first product at the second point in time using the generated first prediction model based on, wherein the generating of the first prediction model includes the daily sales quantity information learning the first prediction model while changing a value of a hyper parameter of the first prediction model by using the first prediction model; and determining a hyper parameter value having the best learning performance as a hyper parameter value of the first prediction model as a result of evaluating learning performance using the daily sales quantity information.

The first prediction model may include a Long Short Term Memory (LSTM) model.

The learning data may include sales quantity information obtained by summing up sales quantities generated in units of a first period from the time of sales start. Also, the first period may be a period of 3 to 10 days.

The method may further include determining whether to use the daily sales quantity information as learning data by determining whether there is no sales quantity information on a specific date among the daily sales quantity information of the first product.

Comparing the daily sales quantity value of the daily sales quantity information with the median value by using a median filter, and as a result of the comparison, selecting a daily sales quantity value at a third time point in which a value deviated from the median value is greater than a threshold value. and replacing the daily sales quantity value of the selected third time point with a value calculated using a median value.

In the step of replacing the value of the daily sales volume at the selected third time point with a value calculated using the median value, the daily sales volume value at the selected third time point is replaced with the average sales volume up to the day before the third time point. and then adding the difference from the replaced value.

The generating of the first prediction model may include generating a plurality of structured candidate prediction models by combining at least some of a plurality of pre-owned AI-based prediction model sequences, the plurality of structured candidate prediction models. inputting and learning a first part of the daily sales quantity information as learning data, evaluating the learning performance of the plurality of structured candidate prediction models using a second part of the daily sales quantity information, and evaluating the evaluation result The method may include selecting a structured candidate prediction model having the best learning performance as the first prediction model.

The plurality of AI-based predictive model sequences may include at least two of an LSTM sequence, a drop out sequence, a dense sequence, and an Ir sequence of an LSTM model.

The plurality of artificial intelligence-based prediction model sequences have hyperparameter values corresponding to each sequence, and the hyperparameters corresponding to each sequence include a first variable associated with the number of memory cells and a first variable associated with the number of input or output attributes. It may include at least one of two variables and a third variable related to the input or output length of time series data.

A method for predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data according to another aspect of the present invention for achieving the above object is to obtain daily sales quantity information of a first product up to a first point in time generating an artificial intelligence-based first prediction model by using the obtained daily sales quantity information as learning data to predict the sales quantity of the first product at the second time point (the second time point is a time point after the first time point) and predicting the sales quantity of the first product at the second time point using the determined first prediction model based on the daily sales quantity information, Generating a predictive model includes generating a structured candidate predictive model by combining at least some of a plurality of artificial intelligence-based predictive model sequences, wherein the plurality of artificial intelligence-based predictive model sequences are LSTM (Long It may include at least two of a short term memory sequence, a drop out sequence, a dense sequence, and an Ir sequence.

2 is a diagram showing an LSTM (Long Short Term Memory) model applicable to a method of predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data according to an embodiment of the present invention by way of example.

Referring to FIG. 2, the device sequentially performs LSTM No. 1 sequence (which can be called a layer), LSTM dropout sequence, LSTM No. 2 sequence, and then a dense sequence. You can create stacked LSTM models. At this time, LSTM sequence #1 takes (1, 6, 1) as an input and outputs (1, 6, 63) as an output. The second sequence, the dropout 1 sequence, receives (1, 6, 63), the output of LSTM sequence 1, as an input and drops out (1, 6, 63) as an output. The lower sequence, LSTM sequence 2, receives (1, 6, 63), which is the output of the dropout 1 sequence, as input and outputs (1, 32). The last dense 1 sequence receives (1, 32) of LSTM 2 sequence as input and outputs (1, 1). Here, dropout is a sequence that excludes an arbitrary input neuron (one-dimensional or two-dimensional) by a specified ratio during learning in order to prevent overfitting. The LSTM sequence is one of the recurrent neural network layers described above, and the dense sequence is a fully-combined layer that connects all input neurons and output neurons.

In addition, in (x,y) of the LSTM model, x is the number of memory cells, y represents the number of input/output attributes in data-dim, and in (x, y, z), z is the input/output of time series data indicates the length.

The device may use an LSTM encoder in which four sequences are mixed in this way. The device does not necessarily have to use an LSTM encoder having the above four sequences, and can find a combination of sequences having optimal learning performance by using a mixture model of various sequences. The device may generate an optimal prediction model by appropriately selecting and arranging the type and sequence of sequences. As described above, the type and order arrangement method of sequences and their setting values may be stored in advance. The device considers combinations of various sequences, but preferably arranges dense sequences so that the last sequence generates a single vector from the input sequence.

3 is an exemplary diagram of a configuration of a device according to an embodiment.

To achieve the above object, an apparatus according to an aspect of the present invention includes a memory for storing daily sales volume information of a first product and instructions related to the operation of a processor, and daily sales volume information of the first product up to a first point in time. is used as learning data to generate an artificial intelligence-based first prediction model to predict the sales quantity of the first product at a second time point (the second time point is a time point after the first time point), and the A processor predicting the sales quantity of the first product at the second point in time using the generated first prediction model based on the daily sales quantity information, wherein the processor: Using the daily sales quantity information, As a result of learning the first prediction model while changing the value of the hyper parameter of the first prediction model and evaluating the learning performance using the daily sales quantity information, the hyper parameter value with the best learning performance may be determined as a value of a hyperparameter of the first predictive model.

An apparatus according to another aspect of the present invention for achieving the above object includes a memory for storing daily sales volume information of a first product and instructions related to the operation of a processor, and daily sales volume information of the first product up to a first point in time. is used as learning data to generate an artificial intelligence-based first prediction model to predict the sales quantity of the first product at a second time point (the second time point is a time point after the first time point), and the A processor for predicting the sales quantity of the first product at the second point in time using the generated first prediction model based on daily sales quantity information, wherein the processor comprises a plurality of artificial intelligence-based predictive model sequences At least some of these are combined to generate a structured candidate prediction model, wherein the plurality of artificial intelligence-based prediction model sequences include a Long Short Term Memory (LSTM) sequence, a drop out sequence, a dense sequence, and Ir sequences.

Device 301 according to one embodiment includes a processor 302 and a memory 303 . The device 301 according to an embodiment may be the above-described server or terminal. The processor may include at least one device described above with reference to FIGS. 1 to 3 or may perform at least one method described above with reference to FIGS. 1 to 3 . The memory 303 may store information related to the above-described method or a program in which the above-described method is implemented. Memory 303 may be volatile memory or non-volatile memory.

The processor 302 may execute a program and control the device 301 . Codes of programs executed by the processor 302 may be stored in the memory 303 . The device 301 may be connected to an external device (eg, a personal computer or network) through an input/output device (not shown) and exchange data.

The embodiments described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, the devices, methods and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA). array), programmable logic units (PLUs), microprocessors, or any other device capable of executing and responding to instructions. A processing device may run an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device. Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

As described above, although the embodiments have been described with limited drawings, those skilled in the art can apply various technical modifications and variations based on the above. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents of the claims are within the scope of the following claims.

Claims (3)

In the method of predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data,
obtaining daily sales quantity information of a first product up to a first point in time;
generating an artificial intelligence-based first prediction model by using the obtained daily sales quantity information as learning data to predict the sales quantity of the first product at a second time point; the second time point is the first is a point in time after a point in time;
predicting the sales quantity of the first product at the second point in time using the generated first prediction model based on the daily sales quantity information; and
selecting one or more recommended products using the generated first prediction model; A method for predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data including According to claim 1,
Generating the first predictive model,
learning the first prediction model while changing a value of a hyper parameter of the first prediction model by using the daily sales quantity information; and
As a result of evaluating learning performance using the daily sales quantity information, determining a hyperparameter value having the best learning performance as a hyperparameter value of the first prediction model,
The generating of the first predictive model includes generating a plurality of structured candidate predictive models by combining at least some of a plurality of previously held artificial intelligence-based predictive model sequences,
At least one of the plurality of structured candidate prediction models includes (i) a first Long Short Term Memory (LSTM) sequence and a second LSTM sequence, wherein input data of the second LSTM sequence is Same as output data - includes, (ii) includes a drop-out sequence for preventing overfitting between the first LSTM sequence and the second LSTM sequence, (iii) dense in the final layer A method for predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data having a form including a dense sequence. According to claim 1,
The step of selecting one or more recommended products,
extracting one or more keywords from the first prediction model;
Scoring each of the plurality of products using the extracted one or more keywords; and
Selecting the one or more recommended products among the plurality of products according to the result of the scoring,
A method for predicting product sales volume and generating product recommendation information using artificial intelligence based on sales data.

Images