KR102644237B1 - vehicle price prediction method using Word2Vec - Google Patents

Abstract

The present invention relates to a method for predicting the price of a vehicle using Word2Vec. The method of the present invention includes a data collection step of extracting and collecting at least one element information from input vehicle image information for predicting the price of a vehicle, and based on the collected element information, word-to-vector corresponding to each element. By using the data processing step of generating Word2Vec and the generated Word2Vec, the price of the vehicle at the time of purchase desired by the user can be predicted by inputting the generated Word2Vec into an artificial intelligence model. there is.

Description

Vehicle price prediction method using Word2Vec {vehicle price prediction method using Word2Vec}

The present invention relates to a method for predicting the price of a vehicle using Word2Vec, and more specifically, to a method for predicting the price of a vehicle using Word2Vec generated to predict the price of a vehicle. It concerns a method of predicting the price of the vehicle at the time of purchase desired by the user by inputting information into an artificial intelligence model.

Recently, with the development of networks such as the Internet, numerous products and goods can be traded through communication networks such as the Internet. In the case of used cars, it has also become possible to trade through networks, and recently, many websites have been created to register used cars for sale and sell them on the network.

However, due to the nature of used car transactions, unlike new cars, used cars do not have a fixed price, and because different prices are determined depending on year, mileage, accident status, color, etc., the prices offered by dealers of used car companies depend on the prices offered by used car dealers. There was an irrationality in the sale price being determined differently.

In general, when buying or selling a used car, you use an offline used car market to find out the market price and trade the car, refer to the online used car market price to purchase a used car online, or P2P (Peer to Peer) transaction, that is, between individuals. Used cars and other automobile transactions are being used. The use of P2P transactions like this is increasing because commissions for used car companies are reduced and transactions are made on equal footing through information exchange between individuals. However, the prices of used cars provided by this method are based on age, etc., and do not accurately reflect the actual condition of the car. Therefore, there is a problem in that standardized price information cannot be provided to buyers.

Recently, in order to supplement this, a method of determining the market price of a used car has been proposed by using accident information, etc. or by additionally reflecting the mileage, etc. However, even in this method, there are many scams such as manipulating the mileage, and the exact condition of parts or operation information of the car. There is a problem that does not reflect the etc.

However, even if the car is of the same model year, the degree of wear of parts due to maintenance varies depending on the car, and the condition of the car varies depending on the driving area or time, so it is necessary to provide consumers with used car price information that reflects the car's attribute information. It has been done.

[Patent Document 1] Korean Patent Publication. 2017.04.12. open.

The present invention was created to solve the above-mentioned problems, and its purpose is to provide a method for predicting the price of a vehicle using Word2Vec.

The objects of the present invention are not limited to the objects mentioned above, and other objects not mentioned can be clearly understood from the description below.

In order to achieve the above objectives, a method for predicting the price of a vehicle using Word2Vec according to an embodiment of the present invention is disclosed. The method includes a data collection step of extracting and collecting at least one element information from the input vehicle image information for predicting the price of a vehicle, and based on the collected element information, word-to-vector (word-to-vector) corresponding to each element information. A data processing step to generate Word2Vec and a step to predict the price of the vehicle at the time of purchase desired by the user by inputting the generated Word2Vec into an artificial intelligence model using the generated Word2Vec. may include.

In addition, according to an embodiment of the present invention, before the vehicle price prediction step, a step of receiving preset element information and purchase time of the vehicle to be determined may be further included.

Additionally, according to an embodiment of the present invention, the artificial intelligence model may be an artificial intelligence model learned by inputting each element information necessary for predicting the price of a vehicle and the depreciation rate information of each element information over time.

In addition, according to an embodiment of the present invention, the artificial intelligence model identifies the input element information, determines the weight of the element information according to the depreciation rate of the element information, and determines the price of the target vehicle based on the weight of the determined element information. can be predicted.

In addition, according to an embodiment of the present invention, when information about the future purchase time is input, the artificial intelligence model readjusts the weights determined according to the depreciation rate of the element information, and based on the readjusted weights, the subject to be judged at the future purchase time You can predict the price of a vehicle.

In addition, according to an embodiment of the present invention, at the time of future purchase, if there is an element whose depreciation rate of element information is more than a preset threshold, the artificial intelligence model readjusts the weight for the element, based on the readjusted weight, You can predict the price of the vehicle you are considering at the time of future purchase.

Additionally, according to an embodiment of the present invention, the element information may include at least one of vehicle manufacturer information, vehicle model information, vehicle year information, vehicle color information, or vehicle mileage information. .

Additionally, according to an embodiment of the present invention, the weight may be reflected at a different depreciation rate depending on each element among the element information.

Additionally, according to an embodiment of the present invention, the weight may be characterized in that, in the case of element information that is related to each other, a depreciation rate of an element with a higher depreciation rate among the depreciation rates of the related element information is applied.

Specific details for achieving the above objectives will become clear by referring to the embodiments described in detail below along with the attached drawings.

However, the present invention is not limited to the embodiments disclosed below, and may be configured in various different forms. In order to ensure that the disclosure of the present invention is complete, those skilled in the art ( It is provided to fully inform those skilled in the art of the invention of the scope of the invention.

According to an embodiment of the present invention, if element information is vectorized using the Word2Vec model, it can easily be numerically different.

Additionally, according to an embodiment of the present invention, element information can be handled numerically, making it possible to predict the price of a vehicle through vector operations on element information.

In addition, according to an embodiment of the present invention, the price of a vehicle can be accurately predicted using simple information, thereby reducing the risk of fraud during vehicle-to-vehicle transactions due to information asymmetry, and enabling transactions due to the accuracy of information. I'm looking forward to it.

The effects of the present invention are not limited to the effects described above, and potential effects expected by the technical features of the present invention can be clearly understood from the description below.

BRIEF DESCRIPTION OF THE DRAWINGS Some of the embodiments are shown in the accompanying drawings so that the above-mentioned features of the present invention may be understood in detail and a more detailed description with reference to the following embodiments. Additionally, similar reference numbers in the drawings are intended to refer to identical or similar functions across various aspects. However, note that the attached drawings only illustrate certain typical embodiments of the present invention and are not to be considered as limiting the scope of the present invention, and other embodiments having the same effect can be fully recognized. Let's do it.
Figure 1 is a flowchart of a method for predicting the price of a vehicle using Word2Vec according to an embodiment of the present invention.
Figure 2 is a diagram showing element information for generating word-to-vector (Word2Vec) according to an embodiment of the present invention.
Figure 3 is a diagram showing a schematic diagram of an artificial intelligence model using Word2Vec according to an embodiment of the present invention.

Since the present invention can be subject to various changes and can have various embodiments, specific embodiments will be illustrated in the drawings and described in detail.

The various features of the invention disclosed in the claims may be better understood by consideration of the drawings and detailed description. The apparatus, method, manufacturing method, and various embodiments disclosed in the specification are provided for illustrative purposes. The disclosed structural and functional features are intended to enable those skilled in the art to specifically implement various embodiments, and are not intended to limit the scope of the invention. The disclosed terms and sentences are intended to easily explain various features of the disclosed invention and are not intended to limit the scope of the invention.

In describing the present invention, if it is determined that a detailed description of related known technology may unnecessarily obscure the gist of the present invention, the detailed description will be omitted.

Hereinafter, a method for predicting the price of a vehicle using Word2Vec according to an embodiment of the present invention will be described.

Figure 1 is a flow chart of a method for predicting the price of a vehicle using Word2Vec according to an embodiment of the present invention, and Figure 2 is element information for generating word-to-vector (Word2Vec) according to an embodiment of the present invention. 3 is a diagram showing a schematic diagram of an artificial intelligence model using Word2Vec according to an embodiment of the present invention.

Referring to Figures 1 to 3, the vehicle price prediction method (S100) according to an embodiment of the present invention collects data by extracting at least one or more element information from the input vehicle image information for predicting the price of the vehicle. Collection step (S101), data processing step (S103) of generating word-to-vectors (Word2Vec) corresponding to each element based on the collected element information, and generation using the generated word-to-vectors (Word2Vec) It may include a step (S105) of predicting the price of the vehicle at the time of purchase desired by the user by inputting the Word2Vec into the artificial intelligence model.

In one embodiment, the data collection step (S101) may be a step of extracting and collecting a plurality of element information from the input vehicle image information to predict the price of the vehicle.

More specifically, element information is information needed to predict the price of a vehicle and may include vehicle manufacturer information, vehicle model information, vehicle year information, vehicle color information, or vehicle mileage information. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In one embodiment, the data processing step (S103) may be a step of generating word-to-vectors (Word2Vec) corresponding to each element based on a plurality of element information collected in the data collection step (S101).

More specifically, the data processing step (S103) is a step of converting text-based car data into numeric vectors. The present invention includes a data processing step (S103), where element information that determines the vehicle price, which is a data format, is converted into vector values and input into an artificial intelligence model to predict the vehicle price.

In other words, the artificial intelligence model includes a data processing step because vectors are the basic data format of neural networks.

Additionally, the data processing step (S103) may generate word-to-vectors (Word2Vec) corresponding to each of vehicle manufacturer information, vehicle model information, vehicle year information, vehicle color information, or vehicle mileage information. By vectorizing each element information, the vectorized element information becomes operable, so the vectorized element information can be processed through computation. By being able to process element information through computation, it becomes possible to predict the price of a vehicle for which there is no accurate data value.

Additionally, by vectorizing element information, organic relationships between related element information can be calculated through an artificial intelligence model.

Here, the Word2Vec model is a word embedding model that includes context and semantic similarity between words in a word expression vector based on predictive learning.

Word2Vec can refer to a vector model used when deep learning techniques are applied to the natural language processing area. In a vector space, when a specific term (subject word) is expressed as a single vector, in order to express the term individually, only the value of the dimension representing the term is '1', and the values of the remaining dimensions are '0', which is sparse. It can be expressed as a vector.

However, in this case, in order to express all the terms that exist in the world, each term has no choice but to be expressed in a space of unknown (or infinite) dimensions, and when judging the similarity with other terms, if the same term exists Problems may arise where meaningful values are derived only from .

To solve this problem, eigenvectors for each term are generated by analyzing all terms occurring in a specific domain, and the eigenvectors can be expressed in a space of finite dimensions (eg, 100 dimensions or 50 dimensions, etc.).

This eigenvector may be referred to as a word-to-vector (Word2Vec), and the value of the word-to-vector (Word2Vec), that is, the value of each dimension, may have an integer or real number value rather than a binary value, and the value of the same dimension The higher the value, the more semantically similar it can be considered. In this way, representing a term as Word2Vec can be referred to as word embedding.

In one embodiment, the method for predicting the price of a vehicle using Word2Vec (S100) includes the step of receiving each element information and purchase point of a preset judgment target vehicle before the vehicle price prediction step (S105). may include.

In one embodiment, the vehicle price prediction step (S105) uses the word-to-vectors (Word2Vec) generated in the data processing step (S103) and inputs the generated word-to-vectors (Word2Vec) into an artificial intelligence model to determine the user's desired result. This may be the step of predicting the price of the vehicle at the time of purchase.

More specifically, in the vehicle price prediction step (S105), the vehicle price can be predicted using each input element information and information about the future purchase time.

In addition, the vehicle price prediction step (S105) is a step of inputting the vectorized word-to-vector of each element information and information about the future purchase point into an artificial intelligence model and predicting the price of the vehicle through the output value. It can be.

Here, an artificial intelligence model is a machine learning algorithm that converts a structured data set of car information into a series of mathematical vectors, which can combine and compare different element vectors to make predictions about car selling price, purchase price, and general demand. there is. Additionally, artificial intelligence models can be created by time series autoregressive and moving average forecasting techniques.

More specifically, the artificial intelligence model may be an artificial intelligence model learned by inputting each element information necessary for predicting the price of a vehicle and the depreciation rate information of each element information over time.

In addition, the artificial intelligence model can identify the input element information, determine the weight of the element information according to the depreciation rate of the element information, and predict the price of the vehicle to be determined based on the weight of the determined element information. Since the depreciation rate varies depending on the vehicle's element information, the price is determined by reflecting this.

In addition, when information about the future purchase time is input, the artificial intelligence model readjusts the weights determined according to the depreciation rate of the element information, and based on the readjusted weights, it can predict the price of the vehicle to be judged at the time of future purchase. This is to re-reflect changes in depreciation rate in vehicle price predictions for a specific point in the future.

For example, the present invention provides not only elements necessary for determining the price of a vehicle (e.g., vehicle color, vehicle mileage, vehicle age, etc.), but also virtual functions (e.g., 3-month sales price, demand score in the last 5 months, etc.) ) can be used to predict the future price of a vehicle. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In addition, the artificial intelligence model, at the time of future purchase, if there is an element whose depreciation rate of element information is more than a preset threshold, readjusts the weight for the element, and based on the readjusted weight, the vehicle subject to the judgment at the time of future purchase price can be predicted. If the depreciation rate exceeds a preset threshold, it becomes an important factor in vehicle price prediction, so this is taken into consideration and reflected.

For example, information such as the age of the vehicle or if the vehicle has a problem with a specific function due to an accident is an important factor in the price of the vehicle, so if it is above the threshold, the depreciation rate may be reflected differently from the normal case. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

Additionally, the weight of the present invention may be characterized in that the depreciation rate is set and reflected differently depending on each element among the element information. Since the depreciation rate varies depending on the element, this is weighted and reflected when predicting the price.

Additionally, in the case of element information that is related to each other, the weight may be characterized by applying a depreciation rate to the element with the highest depreciation rate among the depreciation rates of the related element information. In order to reflect element information related to determining the price of a vehicle, the depreciation rate of the relevant element information can be reflected in predicting the vehicle price by selecting a depreciation rate with a high depreciation rate.

Additionally, Word2Vec has a real number value when vectorized, so a specific value can be derived through operations on vectorized elements. Once a specific value is derived, the price of the vehicle can be predicted analogically using vehicles containing the specific value.

In one embodiment, in the vehicle price prediction step (S105), the price of the vehicle to be judged is predicted by reflecting the correction values (weights) already set for the generated word-to-vectors (Word2Vec) corresponding to each element. You can.

More specifically, the vehicle price prediction step (S105) is performed by a weight assigned to at least one of the element information with respect to the word-to-vectors (Word2Vec) corresponding to each element generated in the data processing step (S103). It can have a determined value.

Here, the preset correction value may be a value determined according to the weight or priority given to at least one of the element information.

Additionally, the weight may be relatively assigned according to the depreciation rate or the price of the elements among the element information.

For example, the weight or priority may be considered for models with high preference, such as a 2010 model, a vehicle with less than 100,000 kilometers traveled, a vehicle with a neutral color, or a limited edition model. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

The artificial intelligence model of the invention extends the capabilities of regular time series prediction models by matching cars with a negligible amount of historical data (cars that are relatively new to the used car market) with cars with a significant amount of historical data (cars with sufficient data). can do.

The above description is merely an illustrative explanation of the technical idea of the present invention, and those skilled in the art will be able to make various changes and modifications without departing from the essential characteristics of the present invention.

Accordingly, the embodiments disclosed in this specification are not intended to limit the technical idea of the present invention, but are for illustrative purposes, and the scope of the present invention is not limited by these embodiments.

The scope of protection of the present invention should be interpreted in accordance with the claims, and all technical ideas within the equivalent scope should be understood to be included in the scope of rights of the present invention.

Claims (9)

In a method for a computing device to predict the price of a vehicle based on Word2Vec,
A data collection step of extracting and collecting at least one element information from the input vehicle image information for predicting the price of a vehicle;
A data processing step of generating word-to-vectors (Word2Vec) corresponding to each element based on the collected element information;
A step of inputting each element information and purchase point of a preset judgment target vehicle; and
A step of predicting the price of the vehicle at the time of purchase desired by the user by inputting the generated Word2Vec into an artificial intelligence model using the generated Word2Vec,
The artificial intelligence model is,
Identifying the inputted element information, determining the weight of the element information according to the depreciation rate of the element information or the price of the elements, and predicting the price of the vehicle to be determined based on the weight of the determined element information,
The artificial intelligence model is,
At the time of future purchase, if there is an element whose depreciation rate of element information is more than a preset threshold, the weight for the corresponding element is readjusted, and based on the readjusted weight, the price of the judgment target vehicle is predicted at the time of future purchase,
The artificial intelligence model is,
Predicting the price of the vehicle subject to judgment by considering non-element information including sales price information within a specific period and demand information for the vehicle within a specific period,
The weight is,
Among the element information, it is set and reflected differently depending on the depreciation rate or the price of the elements, and in the case of element information that is related to determining the price of the vehicle, the depreciation rate of the element with the highest depreciation rate among the depreciation rates of the related element information is applied. ,
Vehicle price prediction method using Word2Vec.
delete According to paragraph 1,
The artificial intelligence model is,
An artificial intelligence model learned by inputting each element information necessary for predicting the price of a vehicle and the depreciation rate information of each element information over time.
Vehicle price prediction method using Word2Vec.
delete delete delete According to paragraph 1,
The above element information is,
Containing at least one of vehicle manufacturer information, vehicle model information, vehicle year information, vehicle color information, or vehicle mileage information,
Vehicle price prediction method using Word2Vec.
delete delete