A method for analyzing automobile competitive products based on k-brown motion clustering

Abstract

The application discloses a kind of car competitive product analysis methods based on k Brown motion clustering, using k Brown motion clustering algorithm for the car competitive product related data that can be divided into k clusters, first randomly select k objects in data as initial cluster center, calculate the distance between each object and each cluster center, each object is assigned to the nearest cluster center, then let the k cluster centers simulate the Brown motion of particle and carry out random motion, after cluster center movement, the objects contained in different categories are updated, the Brown motion of cluster center is repeated until the minimum value of within-cluster sum of squares is found, and the clustering is completed. Since the cluster center moves randomly, the selection of initial cluster center has less influence on the clustering result, and the car competitive products on the market can be more accurately clustered and analyzed by the method, and more accurate market insight and competition strategy for car manufacturers.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of automobile data analysis, and particularly relates to a design of a car competitive product analysis method based on k Brownian motion clustering. BACKGROUND

[0002] The automobile industry is highly competitive, and the market changes rapidly. Different brands and models emerge, and consumer demand is diversified. Under such a background, automobile manufacturers need to keep abreast of market dynamics and the trends of competitors. Competitive product analysis as a market research tool has become a necessary work in the automobile industry. Through comprehensive and detailed analysis of competitive products, automobile manufacturers can provide decision-making reference and improve market competitiveness. Traditional automobile competitive product analysis methods include data statistical analysis, matrix management method, SWOT method, etc. With the continuous development of the field of machine learning, some excellent machine learning algorithms have also been used in the field of automobile competitive product analysis. For example, the k-means algorithm is a classic distance-based clustering algorithm. It takes k as a parameter input, randomly selects k center points, and finally divides n automobile competitive product data into k clusters. In the k clusters, the members in the same cluster have high similarity, and the members in different clusters have high dissimilarity, thereby achieving the purpose of clustering and evaluating analysis of automobile competitive products. The clustering center in the k-means clustering algorithm is determined by calculating the mean of all data object attributes in a cluster. Therefore, the k-means algorithm is usually used to process numerical attributes.

[0003] The k-means algorithm is a very effective clustering method in many practical applications, but the ordinary k-means algorithm has a big drawback, that is, the clustering result will change greatly with the random selection of the initial clustering center, so it cannot guarantee to get a better clustering result, and the accuracy of the final clustering result depends on the selection of the initial clustering center. Therefore, the selection of the initial center point has a great influence on the final clustering result. Selecting appropriate initial center points can speed up the convergence speed of the clustering algorithm and also improve the quality of the clustering result. SUMMARY

[0004] The purpose of the present application is to solve the problem of unstable clustering effect of the existing automobile competitive product analysis method based on the ordinary k-means algorithm, which is greatly affected by the selection of the initial clustering center. The present application proposes an automobile competitive product analysis method based on k Brownian motion clustering. This method has the advantages of high accuracy, good stability, high reliability, and strong real-time performance when clustering and analyzing automobile competitive product data.

[0005] The technical scheme of the present application is as follows: an automobile competitive product analysis method based on k Brownian motion clustering, comprising the following steps:

[0006] S1. Collect relevant data on automotive competitors.

[0007] S2. Preprocess the relevant data of automotive competitors to obtain the sample set D, the number of clusters k, and the maximum number of iterations N.

[0008] S3. Based on the number of clusters k and the maximum number of iterations N, the k-Brownian motion clustering algorithm is used to cluster the sample set D to obtain the clustering results.

[0009] S4. Based on the clustering results, analyze the characteristics and market positioning of each automotive competitor group to obtain the competitor analysis results.

[0010] S5. Based on the results of competitor analysis, formulate corresponding market strategies, product positioning, and improvement plans.

[0011] Furthermore, the relevant data on automotive competitors in step S1 includes vehicle-specific attribute data and vehicle market attribute data.

[0012] Furthermore, the vehicle's own attribute data includes fuel system, fuel type, number of cylinders, peak speed, and vehicle length, width, and height.

[0013] Furthermore, vehicle market attribute data includes vehicle name, vehicle price, and risk assessment level.

[0014] Furthermore, the preprocessing in step S2 includes data cleaning, feature selection, and data normalization.

[0015] Furthermore, step S3 includes the following sub-steps:

[0016] S31. From the sample set D = {x1, x2, ... x...} m Randomly select k samples from the dataset as the initial k cluster center vectors {μ1, μ2, ..., μ3}. k}, and initialize the cluster partition C to Where x i Let μ represent the i-th sample, i = 1, 2, ..., m, where m represents the number of samples in the sample set D. j Let t represent the vector of the j-th cluster center, where j = 1, 2, ..., k, and t = 1, 2, ..., k.

[0017] S32, Calculate sample x i and the cluster center vectors μ j distance d ij :

[0018]

[0019] Where ||·||2 represents the 2-norm.

[0020] S33, classifying the sample x i The minimum distance d is marked ij The corresponding class i The sample x is classified into a cluster i The sample x is classified into a cluster The cluster center is updated

[0021] S34, selecting the maximum Euclidean distance between m samples in the sample set D as the Brownian motion diameter s:

[0022]

[0023] Where x h represents the hth sample, h = 1, 2,..., m.

[0024] S35, a Gaussian random number generation function is used to generate a Gaussian distribution random number gaussRand(0, 1) with an expectation of 0 and a variance of 1.

[0025] S36, a new cluster center vector j is generated according to the Brownian motion diameter s and the Gaussian distribution random number gaussRand(0, 1).

[0026] μ j = μ j-1 + s·gaussRand(0, 1)

[0027] S37, the intra-cluster sum of squares before and after the movement of the cluster center is calculated When the intra-cluster sum of squares after the movement of the cluster center is less than the intra-cluster sum of squares before the movement, the optimal solution of the cluster center is obtained.

[0028] S38, repeating steps S32-S37 until the iteration number is greater than the maximum iteration number N.

[0029] S39, outputting the cluster division C = {C1, C2,..., C k} corresponding to the optimal solution of the cluster center as the clustering result.

[0030] The beneficial effects of the present application are:

[0031] (1) The present application introduces Brownian motion and movement diameter on the basis of traditional clustering methods. Since the cluster center moves randomly, the selection of the initial cluster center has less influence on the clustering result, and the clustering analysis of the automobile competitive data can be more accurate and stable.

[0032] (2) The application provides a high-precision, high-stability and high-reliability automobile competitive product analysis method which can be performed in real time, realizes clustering analysis on collected automobile competitive product data, is more rapid and stable than a traditional analysis method, can timely respond to market demand changes, more accurately analyzes automobile competitive products on the market, and provides automobile manufacturers with more accurate market insight and competitive strategy making basis. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 Fig. 1 shows a flowchart of an automobile competitive product analysis method based on k Brown motion clustering provided by an embodiment of the application. DETAILED DESCRIPTION

[0034] Exemplary embodiments of the application will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described herein are merely exemplary and are intended to illustrate the principles and features of the application, but are not intended to limit the scope of the application.

[0035] An embodiment of the application provides an automobile competitive product analysis method based on k Brown motion clustering, as shown in Fig. 1, which comprises the following steps S1-S5: Figure 1

[0036] S1, collecting relevant data of automobile competitive products.

[0037] In the embodiment of the application, the relevant data of automobile competitive products comprises vehicle self attribute data and vehicle market attribute data. The vehicle self attribute data comprises a fuel system, a fuel type, a cylinder number, a peak speed and a vehicle length, width and height, and the vehicle market attribute data comprises a vehicle name, a vehicle price and a risk assessment level.

[0038] S2, preprocessing the relevant data of automobile competitive products to obtain a sample set D, a clustering cluster number k and a maximum iteration number N.

[0039] In the embodiment of the application, the preprocessing comprises data cleaning, feature selection and data normalization processing. Through variable observation on the original relevant data of automobile competitive products, the data variables are mainly two types of numerical variables and category variables, the numerical variables are subjected to redundant item removal and normalization processing, the category variables are subjected to encoding processing, and then the processed data is integrated into a vector, as shown in Table 1. Finally, a sample set D={x1, x2,...x m} is obtained, wherein the sample set D represents a total of m data, and each data is a vector.

[0040] Table 1

[0041] Fuel type Drive type City mileage Highway mileage …… Car horsepower Car price 1 0 21 27 …… 111 13495 1 0 19 26 …… 154 16500 1 0 24 30 …… 102 13950 …… …… …… …… …… …… ……

[0042] ​S3, based on the cluster number k and the maximum iteration number N, a k Brown motion clustering algorithm is used to cluster the sample set D, and a clustering result is obtained.

[0043] Step S3 includes the following sub-steps S31-S39:

[0044] S31, randomly select k samples from the sample set D={x1, x2,...x m} as initial k cluster center vectors {μ1, μ2,..., μ k}, and initialize the cluster division C to where x i represents the i-th sample, i=1, 2,...m, m represents the number of samples in the sample set D, μ j represents the j-th cluster center vector, j=1, 2,...k, t=1, 2,...k.

[0045] S32, calculate the distance d i between each sample x j and each cluster center vector μ ij :

[0046]

[0047] where ||·||2 represents the 2-norm.

[0048] S33, mark the sample x i as the category λ ij corresponding to the smallest distance d i , and divide the sample x i into the cluster , that is, update

[0049] S34, select the maximum Euclidean distance between the m samples in the sample set D as the Brown motion diameter s:

[0050]

[0051] where x h represents the h-th sample, h=1, 2,...m.

[0052] S35, a Gaussian random number generation function is used to generate a Gaussian distribution random number gaussRand(0, 1) with an expectation of 0 and a variance of 1.

[0053] S36, generate a new cluster center vector μ j according to the Brown motion diameter s and the Gaussian distribution random number gaussRand(0, 1):

[0054] μ j = μj-1 + s * gaussRand(0, 1)

[0055] S37, calculate the within-cluster sum of squares before and after the motion of the cluster center When the within-cluster sum of squares after the motion of the cluster center is less than the within-cluster sum of squares before the motion, the optimal solution of the cluster center is updated.

[0056] S38, repeat steps S32-S37 until the number of iterations is greater than the maximum number of iterations N.

[0057] S39, output the cluster partition C={C1, C2,...C k} corresponding to the optimal solution of the cluster center as the clustering result.

[0058] S4, according to the clustering result, analyze the characteristics and market positioning of each automobile competitive group to obtain a competitive analysis result.

[0059] In the embodiment of the application, automobiles can be roughly divided into five types: mini, general, intermediate, intermediate-high, and high. The four important tags of the cluster are used to analyze the product of the automobile competitors in the data.

[0060] The first type: the driving type is mostly fwd (front-wheel drive), the fuel system is mostly 2bbl (double-cavity fuel system), the vehicle section and the horsepower factor (mainly vehicle length / width / height / horsepower) are lower than the second type, and the price range is concentrated in the low-middle price range. According to this category, the automobiles in this category mostly use front-wheel drive, traditional double-cavity fuel system, and ordinary vehicle spaciousness and horsepower, and the prices are mostly distributed in the low-middle price range. Therefore, this category belongs to "economy car", and the product positioning belongs to low-end economy car. The brands with more vehicles in this category are Toyota, Mitsubishi, Honda, Subaru, Nissan, Mazda, and Volkswagen.

[0061] The second type: the driving type is mostly rwd (rear-wheel drive), the fuel system is mostly mpfi (multi-point fuel system), the vehicle section and the horsepower factor (mainly vehicle length / width / height / horsepower) are higher than the first type, and the price range is concentrated in the middle-high price range. According to this category, the automobiles in this category mostly use rear-wheel drive, multi-point fuel system, and excellent vehicle spaciousness and horsepower, and the prices are mostly distributed in the middle-high price range. Therefore, this category belongs to "luxury car", and the product positioning belongs to middle-high-end luxury car / racing car. The brands with more vehicles in this category are BMW, Porsche, Jaguar, and Alpha Romeo.

[0062] S5, based on the competitive analysis result, formulate corresponding market strategy, product positioning, and improvement scheme.

[0063] Those skilled in the art will appreciate that the embodiments described herein are presented for purposes of illustration and that the inventive principles are not limited to these particular embodiments. Other variations and modifications can be made to the embodiments without departing from the spirit and scope of the inventive principles.

Claims

1. A method for automotive competitive analysis based on k Brownian motion clustering, characterized by, The method comprises the following steps: S1, collecting relevant data of automobile competitors; S2, preprocessing the relevant data of automobile competitors to obtain a sample set D, a clustering cluster number k, and a maximum iteration number N; S3, based on the number of clusters k and the maximum number of iterations N, using k Brownian motion clustering algorithm to cluster the sample set D, to get the clustering results; S31, randomly select k samples from the sample set D={x1, x2,...x m} as the initial k cluster center vectors {μ1, μ2,..., μ k}, and initialize the cluster division C as Where x i represents the i-th sample, i=1, 2,...m, m represents the number of samples in the sample set D, μ j represents the j-th cluster center vector, j=1, 2,...k, t=1, 2,...k; S32, calculate sample x i and each cluster center vector μ j distance d ij : Wherein ||·||2 represents 2-norm; S33, classifying the sample x i the distance d ij the corresponding class λ i the sample x i into a cluster i.e. updating S34, selecting the maximum Euclidean distance between m samples in the sample set D as the Brownian motion diameter s: where x h represents the hth sample, h = 1, 2,... m; S35, generating a Gaussian distribution random number gaussRand(0,1) with an expectation of 0 and a variance of 1 by using a Gaussian random number generation function; S36, generating a new cluster center vector μ according to the Brownian motion diameter s and a Gaussian distribution random number gaussRand(0, 1) j : μ j = μ j-1 + s·gaussRand(0,1) S37、calculating the within-cluster sum of squares before and after the motion of the cluster center updating the optimal solution of the cluster center when the within-cluster sum of squares after the motion of the cluster center is less than the within-cluster sum of squares before the motion of the cluster center S38, repeating steps S32-S37 until the iteration number is greater than the maximum iteration number N; S39, output the cluster center optimal solution corresponding to the cluster partition C = {C1, C2,... C k} as the clustering result; S4, analyzing the characteristics and market positioning of each automobile competitor group according to the clustering results to obtain competitor analysis results; S5, formulating corresponding market strategies, product positioning, and improvement schemes based on the competitor analysis results.

2. The automotive competitor analysis method according to claim 1, characterized by, The relevant data of automobile competitors in step S1 includes vehicle self attribute data and vehicle market attribute data.

3. The automotive competitor analysis method according to claim 2, characterized by, The vehicle self attribute data includes fuel system, fuel type, cylinder number, peak speed, and vehicle length, width, and height.

4. The automotive competitive analysis method according to claim 2, characterized by, The vehicle market attribute data includes vehicle name, vehicle price, and risk assessment level.

5. The automotive competitor analysis method according to claim 1, characterized by, The preprocessing in step S2 includes data cleaning, feature selection, and data normalization processing.

Images