Consumption decline prediction method and system for gas station customers

Abstract

The invention relates to the technical field of gas station big data analysis, and particularly discloses a consumption decline prediction method and system for a gas station customer, and the method comprises the steps: obtaining the transaction data of the gas station customer; calculating the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price arrival rate of each customer; pCA dimension reduction processing is carried out, then GMM clustering is carried out, and a clustering result is obtained; screening out a to-be-analyzed customer group; the expected refueling interval day number, the interval date standard deviation and the time interval from the last refueling time to the current time of each customer are calculated; and judging whether the customer is a consumption decline customer. The customer group is subdivided based on the improved RFM model and the Gaussian mixture model, and the time sequence analysis, the statistics principle and the data analysis technology are combined to dynamically monitor the consumption behavior change of the customer so as to predict the decline sign and the potential loss risk of the customer and provide a timely early warning signal.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis of gas stations, and in particular to a consumption decline prediction method and system for gas station customers. Background Art

[0002] In high-frequency consumption scenarios such as gas stations, understanding customer consumption habits is crucial to improving customer satisfaction and optimizing operations. Currently, the industry mainly uses the following methods to analyze customer behavior: Traditional statistical aggregation: This method is usually based on simple counting and summing calculations, such as monthly sales, average purchase frequency, etc. It can provide some basic information, but it is difficult to reveal deep behavioral patterns.

[0003] Rule-driven approach: By setting a series of business rules, for example, if a customer does not consume for three consecutive months, the customer is considered lost. However, this approach lacks flexibility and cannot adapt to complex market changes and individual differences.

[0004] Static segmentation based on the RFM model: The RFM (Recency, Frequency, Monetary) model is a commonly used customer segmentation method that segments customers based on their most recent transaction time, transaction frequency, and transaction amount. Although the RFM model can effectively distinguish customer groups of different values, it is essentially a static analysis tool that fails to consider dynamic changes in time series.

[0005] Simple time series forecasting: Some companies use time series models such as ARIMA to predict future sales trends. This type of method is suitable for data with obvious periodicity, but has limited effect on predicting consumer behavior under nonlinear and multi-factor influences.

[0006] Although the above methods have helped companies understand their customers to a certain extent, they still have the following shortcomings: Lack of personalized insights: Traditional methods can only provide data analysis results at the overall or group level, making it difficult to provide customized marketing strategies and service recommendations for individual customers.

[0007] Ignoring the impact of the time dimension: Most existing analysis methods fail to fully consider the dynamic changes in time series, resulting in inaccurate predictions of customer behavior. For example, relying solely on the RFM model cannot capture the possibility of a customer suddenly increasing or decreasing consumption in the short term.

[0008] Weak ability to respond to market changes: With increasingly fierce market competition and rapid technological development, consumer needs and preferences are constantly changing. Existing analytical methods are often fixed and difficult to quickly respond to new market trends and customer needs.

[0009] Lack of early warning mechanism: Most existing analytical tools focus on post-event summaries rather than pre-event predictions. There is a lack of an effective early warning system to identify customers whose consumption may decline in advance, thus missing out on recovery opportunities.

[0010] In summary, although existing customer analysis methods can help companies better understand their customer base in some aspects, there is still much room for improvement in terms of personalization, dynamism, and market response speed. Summary of the invention

[0011] The present invention aims to solve the above problems. To this end, the present invention provides a consumption decline prediction method and system for gas station customers, which segment customer groups based on an improved RFM model (Recency, Frequency, Monetary) and Gaussian Mixture Model (GMM), and dynamically monitor changes in customer consumption behavior in combination with time series analysis, statistical principles, and data analysis technology to predict customer decline signs and potential loss risks, and provide timely early warning signals.

[0012] The present invention provides a consumption decline prediction method for gas station customers, and the technical solution adopted is as follows: Obtain gas station customer transaction data; According to the gas station customer transaction data, the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price implementation rate of each customer are calculated; Perform PCA dimensionality reduction processing on the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price availability rate of all customers, and then perform GMM clustering to obtain the clustering results; Filtering the customer group to be analyzed from the clustering results; Calculate the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present for each customer based on the gas station customer transaction data of the customer group to be analyzed; Whether the customer is a consumption decline customer is determined based on the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present.

[0013] Furthermore, the calculation formula for the expected refueling interval days EBI is: in, Indicates the amount of fuel last refueled. It is the median of all unit oil intervals.

[0014] Furthermore, the calculation formula for the unit oil product interval is: in, represents the interval of unit oil products for the i-th customer, represents the transaction interval between the i-th and i+1-th refueling, Represents the amount of fuel for the i-th refueling.

[0015] Furthermore, the criteria for judging customers with consumption decline are: like , then the customer is a customer with severe consumption decline; like , then the customer is a general consumption decline customer; in, represents the standard deviation of the interval date, Indicates the time interval from the last refueling to the present time.

[0016] Furthermore, the gas station customer transaction data includes customer name, timestamp, amount of fuel consumed, consumption amount and discount amount.

[0017] Furthermore, the calculation formula for the time interval R from the last refueling to the current time is: in, Indicates the largest transaction timestamp among all transactions of the customer. Indicates the current date; The calculation formula of price availability rate P is: in, Indicates the actual amount received by the customer. Indicates the amount receivable from the customer.

[0018] Furthermore, the process of screening out the customer groups to be analyzed from the clustering results is as follows: according to the distribution of the time interval from the last refueling to the present, the total number of transactions, the total consumption amount and the price availability rate of each customer group in the clustering results, one or more customer groups are selected as the customer groups to be analyzed.

[0019] Furthermore, the selection criteria for the customer group to be analyzed are: short time interval between the last refueling and the current time, low total transaction number, high total consumption amount and low price availability rate.

[0020] Furthermore, customers who do not meet any of the following criteria are eliminated from the customer group to be analyzed: The number of blank days for the last oil consumption is less than the threshold number of days; The minimum number of consumptions by the customer in the period before the last consumption is less than the number threshold; The member account balance is less than the balance threshold.

[0021] The present invention also provides a consumption decline prediction system for gas station customers, which adopts the following technical solution: comprising: a data acquisition module, an RFM-P model analysis module, a cluster analysis module, a customer group screening module to be analyzed, and a consumption decline customer judgment module connected in sequence, wherein the consumption decline customer judgment module is also connected to the data acquisition module; The data acquisition module is used to obtain gas station customer transaction data; The RFM-P model analysis module is used to calculate the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price implementation rate of each customer based on the customer transaction data of the gas station; The cluster analysis module is used to perform PCA dimensionality reduction processing on the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price availability rate of all customers, and then perform GMM clustering to obtain clustering results; The to-be-analyzed customer group screening module is used to screen the to-be-analyzed customer group from the clustering results; The consumption decline customer judgment module is used to calculate the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present for each customer based on the gas station customer transaction data of the customer group to be analyzed; and judge whether the customer is a consumption decline customer based on the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present.

[0022] The above one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: The present invention designs the RFM-P model, which focuses on the price availability rate in the model and combines it with the Gaussian mixture model to achieve deep customer segmentation, accurately identify the most valuable customer groups, and reveal the subtle differences between different customer groups through further cluster analysis. It provides companies with unprecedented insights and helps to better meet the unique needs of each market segment.

[0023] The present invention conducts in-depth mining of the clustering results of the RFM-P model data, screens out the customer group to be analyzed, and calculates the expected number of days between refueling by performing time series analysis on the gas station customer transaction data of the customers in this group, estimates the customer's future purchasing tendency, and identifies the customers with consumption decline. The present invention can identify high-value customers who may be lost at an early stage, which helps enterprises implement preventive measures to maintain customer loyalty and prevent customer consumption decline.

[0024] The present invention can effectively identify customers who are about to have abnormal consumption. When customers show obvious signs of abnormal consumption, they will be judged as customers with declining consumption, which helps enterprises take necessary measures in time, slow down or even block the development of customers into lost customers, optimize their life cycle types, and improve customer activity.

[0025] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0027] Figure 1 It is a flow chart of the method provided by the present invention.

[0028] Figure 2 It is a cumulative variance contribution rate diagram of the method provided by the present invention.

[0029] Figure 3 is a scree plot of the method provided by the present invention.

[0030] Figure 4 It is a scatter plot of the data structure of the method provided by the present invention.

[0031] Figure 5 It is a GMM clustering result diagram of the method provided by the present invention.

[0032] Figure 6 It is a distribution diagram of characteristic values ​​of each group of the method provided by the present invention.

[0033] Figure 7 It is a structural block diagram of the system provided by the present invention.

[0034] Reference numerals: 1. Data collection module; 2. RFM-P model analysis module; 3. Cluster analysis module; 4. Customer group screening module for analysis; 5. Consumption decline customer judgment module. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present invention. The following embodiments are used to illustrate the present invention, but cannot be used to limit the scope of the present invention.

[0036] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the embodiment of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0037] Combine the following Figures 1 to 7 The present invention is further described in detail, and a consumption decline prediction method and system for gas station customers of the present invention are described: In this embodiment, Figure 1 As shown, a consumption decline prediction method for gas station customers is provided, comprising the following steps: Step 1: Obtain gas station customer transaction data.

[0038] The gas station customer transaction data includes information such as customer name, timestamp, fuel type, fuel volume (liters), consumption amount, discount amount, etc.

[0039] Gas station customer transaction data comes from the gas station's transaction record files. Conventional gas station transaction record files cannot be directly used as gas station customer transaction data, and generally need to be cleaned. Data cleaning includes: deleting missing values; limiting the range of each customer's oil volume according to the box plot method to exclude abnormal data points; and screening valid transaction records within a specific time period.

[0040] In addition, in order to focus on the behavior of a specific group, the cleaned data can also be sliced, for example, to filter out the target subset from the data based on custom conditions such as date range, upper and lower limits of fuel volume, station name or code, etc. All gas station customer transaction data or the data in the target subset can be analyzed later.

[0041] Step 2: Based on the gas station customer transaction data, calculate the time interval from the last refueling to the present, the total number of transactions, the total consumption amount and the price implementation rate for each customer.

[0042] The traditional RFM model focuses on three dimensions: the time of the last purchase (Recency), the frequency of purchase (Frequency), and the total amount of consumption (Monetary). However, these indicators cannot fully capture the customer's behavior patterns, especially for those price-sensitive customers or customer groups that are more responsive to promotions. Therefore, this method introduces the price fulfillment rate to supplement the deficiencies of the traditional RFM customer classification model, enhance the accuracy of customer segmentation, and record the improved classification model as the RFM-P model. The RFM-P model focuses on four dimensions: the time interval from the last refueling to the current time, the total number of transactions, the total amount of consumption, and the price fulfillment rate.

[0043] The calculation formula for the time interval R from the last refueling to the current time is: in, Indicates the largest transaction timestamp among all transactions of the customer. Indicates the current date.

[0044] Total transaction times F: counts the total transaction times of each customer.

[0045] Total consumption amount M: summarizes the total consumption amount of each customer.

[0046] The calculation formula of price availability rate P is: in, Indicates the actual amount received by the customer. Indicates the amount receivable from the customer. The difference between the amount receivable and the amount actually collected is the discount amount.

[0047] Data standardization: Use MinMaxScaler to normalize the customer RFM-P values ​​to ensure that data of different magnitudes can be treated fairly. This facilitates subsequent customer classification based on their RFM-P values.

[0048] For each feature value x in R, F, M, MinMaxScaler applies the following formula to calculate the normalized value : Among them, min(x) is the smallest value of the feature in the entire data set, and max(x) is the largest value of the feature in the entire data set.

[0049] Step 3: Perform PCA dimensionality reduction processing on the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price implementation rate of all customers, and then perform GMM clustering to obtain the clustering results.

[0050] PCA Dimensionality Reduction: PCA (Principal Component Analysis) is a widely used dimensionality reduction technique that transforms a set of variables that may be correlated into a set of linearly independent variables through linear transformation. These new variables are called "principal components". These new variables are linear combinations of the original variables and are sorted by variance, so that the first principal component has the largest variance, and so on. In this way, a few principal components can be selected to approximate the original data, thereby reducing the data dimension, speeding up the calculation, and reducing the risk of overfitting.

[0051] Let the original data matrix be X, , where n is the number of samples and p is the number of features. In this example, p=4. is the real number field. The goal of PCA is to find a projection matrix W, , so that , Y represents the data matrix after dimensionality reduction, and the columns (principal components) of Y are mutually orthogonal and have the maximum variance. The specific steps are as follows: 1. Centering: Subtract the mean of each column feature to get a new data matrix .

[0052] in, represents the element in the i-th row and j-th column of the original data matrix X, represents the average value of the jth column in the original data matrix X, Represents the new data matrix The element at row i and column j in .

[0053] 2. Covariance matrix calculation: Calculate the covariance matrix C.

[0054] in, Represents the new data matrix The transpose of .

[0055] 3. Eigenvalue decomposition: solve the eigenvalues ​​and corresponding eigenvectors of the covariance matrix C, and sort them by eigenvalue size.

[0056] 4. Select principal components: Select the eigenvectors corresponding to the first k eigenvalues ​​to form the projection matrix W to achieve data dimensionality reduction.

[0057] 5. Projection: Project the original data into a low-dimensional space.

[0058] Among them, Y represents the data matrix after dimensionality reduction.

[0059] In this embodiment, after data cleaning of the actual transaction record file of gas station A, the gas station customer transaction data is obtained, and then the corresponding standardized RFM-P data is calculated.

[0060] The process of PCA dimensionality reduction processing on the above-mentioned standardized RFM-P data in this embodiment is as follows: first, the standardized RFM-P data (the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price availability rate) is PCA transformed to calculate the cumulative variance contribution rate, such as Figure 2 As shown, and a scree plot is drawn, such as Figure 3 As shown in , the combination of the two determines that the number of principal components that need to be retained is 2. Figure 2 and Figure 3 It can be seen that when the number of principal components is 2, 99.54% of the variance is explained. This example selects the first two principal components as the basis for subsequent analysis and draws a scatter plot of the data structure after PCA dimensionality reduction of the RFM-P model data, as shown in Figure 4 As shown, Figure 4 The principal components A and B in are the two principal components that need to be retained.

[0061] GMM clustering assumes that the observed data is a mixture of multiple Gaussian distributions, each of which represents a potential customer group. The model parameters include the mean covariance matrix and prior probability of each Gaussian component. During the training process, the expectation maximization (EM) algorithm is used to iteratively update these parameters until convergence. Ultimately, for each customer, GMM will give the probability value of belonging to each cluster, thereby achieving soft assignment.

[0062] In this embodiment, the RFM-P model data after PCA dimension reduction is clustered by GMM, and the GMM clustering result is as follows: Figure 5 As shown in Figure 1, customers in the RFM-P model are divided into four groups. The distribution of the characteristic values ​​of the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount, and the price fulfillment rate of each group is shown in Figure 1. Figure 6 As shown in (a)-(d). Figure 6 The horizontal axes in (a)-(d) are the group numbers, i.e., from left to right, they are group 0, group 1, group 2, and group 3. The order of the groups is the same as Figure 5 The four parts of the GMM clustering results correspond to Group 0, Group 1, Group 2, and Group 3 respectively. Figure 5 The four parts from left to right.

[0063] according to Figure 5 The distribution of eigenvalues ​​can be used to define the customer groups in the RFM-P model: The active new customer group (Group 0) has recently purchased, has a low purchase frequency, consumes less, and has a low price fulfillment rate.

[0064] The high-value decline group (Group 1) has a long time to purchase, a high purchase frequency, a small amount of consumption, and a high price fulfillment rate.

[0065] The potential large-amount group (Group 2) has recently made purchases, has a low purchase frequency, a high consumption amount, and a low price fulfillment rate.

[0066] The dormant low-price group (Group 3) had a long time ago as the last purchase, a low purchase frequency, a small amount of consumption, and a low price fulfillment rate.

[0067] In order to verify the difference between the RFM-P model and the RFM model, in this embodiment, the data of the standardized RFM model corresponding to the gas station customer transaction data of gas station A is calculated. Then the same PCA dimensionality reduction and GMM clustering processing methods are used to process the data of the RFM model. Through PCA dimensionality reduction processing, the number of principal components that need to be retained in the RFM model is determined. The results show that when the number of principal components is 2, 99.98% of the variance is explained. After GMM clustering, the average silhouette coefficient of the clustering results obtained from the RFM model data is 0.15. The average silhouette coefficient of the RFM-P model data is 0.59, which reflects that after the introduction of the price implementation rate, the customer grouping effect is better.

[0068] Step 4: Filter out the customer group to be analyzed from the clustering results.

[0069] Based on the clustering results, a specific group is selected for analysis. The process of selecting the customer groups to be analyzed from the clustering results is as follows: according to the distribution of the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price fulfillment rate of each customer group in the clustering results, one or more customer groups are selected as the customer groups to be analyzed. Generally speaking, one customer group is selected each time for subsequent analysis.

[0070] In this embodiment, the selection criteria of the customer group to be analyzed are: short time interval between the last refueling and the current time, low total transaction number, high total consumption amount and low price availability. By making a comprehensive comparison of the four groups, group 2 is determined as the customer group to be analyzed.

[0071] The customer groups to be analyzed can also be checked for activity to determine whether the customers remain active, remove active customers, and screen out inactive customers for subsequent analysis. Specifically: remove customers who do not meet any of the following criteria from the customer groups to be analyzed; Standard 1: The number of blank days for the last oil consumption is less than the threshold number of days; Standard 2: The minimum number of consumptions by the customer in the period before the last consumption is less than the number threshold; Criterion 3: The member account balance is less than the balance threshold.

[0072] In this embodiment, the day threshold is 90 days; the period before the last consumption is 6 months, the number threshold is 4 times; and the balance threshold is 200 yuan.

[0073] Step 5: Based on the gas station customer transaction data of the customer group to be analyzed, calculate the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present for each customer.

[0074] In this step, time series analysis is performed on the gas station customer transaction data of the customer group to be analyzed. Time series analysis: Using the simple moving average method, the time difference between each transaction and the next transaction is calculated to obtain the transaction interval time (days) between each refueling of the customer. The amount of oil consumed in the time interval between two refuelings is the amount of oil consumed by the customer last time, so as to further calculate the transaction interval days corresponding to each liter of oil, that is, how many days each liter of oil can support the customer's use, which is defined as the unit oil interval, to determine the regularity and frequency of the customer's refueling behavior, and then calculate the customer's next consumption time interval based on the calculated unit oil interval value and the last refueling oil volume value, as the expected refueling interval days for predicting customer consumption behavior.

[0075] The calculation formula for unit oil interval UFI is: in, represents the interval of unit oil products for the i-th customer, It represents the transaction interval between the i-th and i+1-th refueling, in days. Represents the amount of fuel added for the i-th time, in liters.

[0076] The calculation formula for the expected refueling interval (EBI) is: in, Indicates the amount of fuel last refueled. It is the median of all unit oil intervals.

[0077] The standard deviation of interval dates SDID is the standard deviation of the interval time (days) of all customer transactions. The calculation formula is: Among them, m represents the total number of refueling times, Indicates the average time between transactions.

[0078] Step 6: Determine whether the customer is a consumption decline customer based on the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present.

[0079] The warning level determination is used to determine whether the customer is in a general warning or severe warning state, based on the relationship between the time interval from the customer's last refueling to the present and the expected basic interval date and the standard deviation of the interval date. Using the 3sigma criterion, that is, when the time interval between the last consumption date and the modeling date exceeds the sum of the expected number of refueling interval days and the standard deviation of the interval date, it is defined as a consumption recession warning of varying degrees.

[0080] Warning level classification: According to different warning conditions, customers are divided into two categories: general warning and severe warning (general warning is two times, severe warning is three times), so as to formulate targeted marketing strategies.

[0081] The criteria for judging customers with consumption recession are: like , then the customer is a customer with severe consumption decline; like , the customer is a general consumption decline customer.

[0082] Design customer marketing strategies for customers experiencing consumption recession, including: Preference identification: Based on the fuel products consumed by the customer in a specific time period (such as 6 months before the last consumption) and the gas stations they frequently visit, we count the locations of each member's most frequently visited gas stations and the types of fuel products they most commonly use, so as to better understand their geographical location and product preferences, and to distribute exclusive coupons for limited gas stations or fuel product numbers to customers.

[0083] Coupon customization: Design personalized discount plans based on the customer's consumption history and personal characteristics, especially for those customers who are marked as severely declining consumers. In this example, the potential large-amount customers selected generally have a low purchase frequency, but a large amount of each purchase, showing a large consumption potential. However, compared with other customer groups, they have a lower price completion rate, so they may be sensitive to prices. Therefore, we choose to distribute exclusive discount coupons to this type of customers to achieve the purpose of increasing their purchase frequency.

[0084] Execute feedback loops: track the effectiveness of implemented marketing measures, collect customer response data, and continuously optimize marketing strategies based on the customer participation rate of each marketing campaign.

[0085] In the process of practical application, 82% of the customers predicted to be in consumption decline had intermittent consumption behavior in the following three months, and the possibility of consumption decline was high. In addition, through effective marketing intervention, the consumption behavior of most customers has been improved. Especially among the customer groups with severe consumption decline, the 95# gasoline customer group responded to the marketing activities better than the 92# gasoline customer group. In general, this method improves the customer retention rate and activity of sales companies, and maintains and expands the stable customer group.

[0086] This embodiment also provides a consumption decline prediction system for gas station customers, such as Figure 7 As shown, the adopted technical solution is as follows: it includes: a data acquisition module 1, an RFM-P model analysis module 2, a cluster analysis module 3, a customer group screening module to be analyzed 4 and a consumption decline customer judgment module 5 which are connected in sequence, and the consumption decline customer judgment module 5 is also connected to the data acquisition module 1.

[0087] The data acquisition module 1 is used to obtain gas station customer transaction data; The RFM-P model analysis module 2 is used to calculate the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price implementation rate of each customer based on the customer transaction data of the gas station; The cluster analysis module 3 is used to perform PCA dimensionality reduction processing on the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price availability rate of all customers, and then perform GMM clustering to obtain clustering results; The to-be-analyzed customer group screening module 4 is used to screen the to-be-analyzed customer group from the clustering results; The consumption decline customer judgment module 5 is used to calculate the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present for each customer based on the gas station customer transaction data of the customer group to be analyzed; and judge whether the customer is a consumption decline customer based on the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present.

[0088] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A consumption decline prediction method for gas station customers, characterized in that: include: Obtain gas station customer transaction data; According to the gas station customer transaction data, the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price implementation rate of each customer are calculated; Perform PCA dimensionality reduction processing on the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price availability rate of all customers, and then perform GMM clustering to obtain the clustering results; Filtering the customer group to be analyzed from the clustering results; Calculate the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present for each customer based on the gas station customer transaction data of the customer group to be analyzed; Whether the customer is a consumption decline customer is determined based on the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present.

2. A consumption decline prediction method for gas station customers according to claim 1, characterized in that: The calculation formula for the expected refueling interval (EBI) is: in, Indicates the amount of fuel last refueled. It is the median of all unit oil intervals.

3. A consumption decline prediction method for gas station customers according to claim 2, characterized in that: The calculation formula for unit oil product interval is: in, represents the interval of unit oil products for the i-th customer, represents the transaction interval between the i-th and i+1-th refueling, Represents the amount of fuel for the i-th refueling.

4. A consumption decline prediction method for gas station customers according to claim 2 or 3, characterized in that: The criteria for judging customers with consumption recession are: like , the customer is a customer with severe consumption decline; like , then the customer is a general consumption decline customer; in, represents the standard deviation of the interval date, Indicates the time interval from the last refueling to the present time.

5. A consumption decline prediction method for gas station customers according to claim 1, characterized in that: The gas station customer transaction data includes customer name, timestamp, amount of fuel refueled, consumption amount and discount amount.

6. A consumption decline prediction method for gas station customers according to claim 1 or 5, characterized in that: The calculation formula for the time interval R from the last refueling to the current time is: in, Indicates the largest transaction timestamp among all transactions of the customer. Indicates the current date; The calculation formula of price availability rate P is: in, Indicates the actual amount received by the customer. Indicates the amount receivable from the customer.

7. A consumption decline prediction method for gas station customers according to claim 1, characterized in that: The process of selecting the customer groups to be analyzed from the clustering results is as follows: according to the distribution of the time interval from the last refueling to the present, the total number of transactions, the total consumption amount and the price availability rate of each customer group in the clustering results, one or more customer groups are selected as the customer groups to be analyzed.

8. A consumption decline prediction method for gas station customers according to claim 7, characterized in that: The selection criteria for the customer group to be analyzed are: short time interval between the last refueling and the current time, low total number of transactions, high total consumption amount and low price availability.

9. A consumption decline prediction method for gas station customers according to claim 7 or 8, characterized in that: Eliminate customers who do not meet any of the following criteria from the customer group to be analyzed: The number of blank days for the last oil consumption is less than the threshold number of days; The minimum number of consumptions by the customer in the period before the last consumption is less than the number threshold; The member account balance is less than the balance threshold.

10. A consumption decline prediction system for gas station customers, characterized in that: A method for predicting consumption decline of gas station customers according to any one of claims 1 to 9, comprising: a data acquisition module, an RFM-P model analysis module, a cluster analysis module, a customer group screening module to be analyzed and a consumption decline customer judgment module connected in sequence, wherein the consumption decline customer judgment module is also connected to the data acquisition module; The data acquisition module is used to obtain gas station customer transaction data; The RFM-P model analysis module is used to calculate the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price implementation rate of each customer based on the customer transaction data of the gas station; The cluster analysis module is used to perform PCA dimensionality reduction processing on the time interval from the last refueling to the current time, the total number of transactions, the total consumption amount and the price availability rate of all customers, and then perform GMM clustering to obtain clustering results; The to-be-analyzed customer group screening module is used to screen the to-be-analyzed customer group from the clustering results; The consumption decline customer judgment module is used to calculate the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present for each customer based on the gas station customer transaction data of the customer group to be analyzed; and judge whether the customer is a consumption decline customer based on the expected number of days between refueling, the standard deviation of the interval dates, and the time interval from the last refueling to the present.

Images