Analysis method and system based on tobacco flow gravity model

Abstract

The application relates to the technical field of data processing analysis, in particular to an analysis method and system based on a tobacco flow gravity model. The method comprises the following steps: obtaining sales data, production data, monthly flow back data and geographic auxiliary data of a target tobacco terminal; inputting the sales data and the production data into a target flow gravity model to obtain a terminal gravity value; performing abnormality detection on the terminal gravity value to obtain a target abnormal gravity value; performing prediction on the monthly flow back data and the geographic auxiliary data through a hybrid prediction model to obtain a predicted flow amount; and performing abnormal flow analysis on the target abnormal gravity value and the predicted flow amount to obtain an abnormal flow area range. The application can consider the mutual influence between terminals, adapt to market dynamic changes, and continuously optimize itself to adapt to complex and changeable market environments.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing analysis, and in particular to an analysis method and system based on a tobacco flow gravitational model. BACKGROUND

[0002] In the related art, with the rapid development of the tobacco industry and the increasingly fierce market competition, the illegal tobacco circulation problem of tobacco cross-regional circulation and sales is increasingly prominent, which seriously affects the healthy development of the industry. Existing supervision methods include methods such as relying on manual patrol, abnormal detection based on sales data, flow direction analysis based on logistics data, or abnormal circulation detection by applying machine learning algorithms. However, the existing supervision methods often have the following problems: there is a lag, it is difficult to discover and prevent illegal circulation behavior in a timely manner; only single-dimensional data is concerned, and the comprehensive consideration of multi-dimensional factors of the market is lacking; static supervision methods are difficult to adapt to the rapidly changing market environment and the constantly evolving illegal circulation means; the mutual influence between terminals and the spatial relationship of the market are ignored, and the overall situation of the market cannot be accurately described; there is a lack of effective abnormal processing and model optimization mechanism, and it is difficult to continuously improve the accuracy and reliability of detection.

[0003] In summary, the technical problems in the related art need to be improved. SUMMARY

[0004] The main purpose of the embodiments of the present application is to propose an analysis method and system based on a tobacco flow gravitational model, which can consider the mutual influence between terminals, adapt to market dynamics, and continuously optimize itself to adapt to the complex and changing market environment.

[0005] To achieve the above-mentioned purpose, one aspect of the embodiments of the present application proposes an analysis method based on a tobacco flow gravitational model, which comprises:

[0006] obtaining sales data, production data, monthly flow back data and geographic auxiliary data of a target tobacco terminal;

[0007] inputting the sales data and the production data into a target flow gravitational model to obtain a terminal gravitational value;

[0008] performing abnormal detection on the terminal gravitational value to obtain a target abnormal gravitational value;

[0009] predicting the monthly flow back data and the geographic auxiliary data by a hybrid prediction model to obtain a predicted flow amount;

[0010] performing abnormal flow analysis on the target abnormal gravitational value and the predicted flow amount to obtain an abnormal flow area range.

[0011] In some embodiments, the method for constructing the target flow gravity model comprises the following steps:

[0012] Obtaining the total historical sales and the total historical production of the region to be detected;

[0013] Processing the total historical sales and the total historical production by the initial flow gravity model to obtain terminal gravity values;

[0014] Correcting the initial flow gravity model by a particle model to obtain the target flow gravity model.

[0015] In some embodiments, the method for correcting the initial flow gravity model by the particle model to obtain the target flow gravity model comprises the following steps:

[0016] Constructing a plurality of particles with unit mass around the target tobacco terminal, the particles being distributed between the target tobacco terminal and other terminals;

[0017] Obtaining the particle distance between two adjacent particles;

[0018] Obtaining the particle gravity of the particle farther from the target tobacco terminal in the particle distance;

[0019] Calculating the ratio of the particle gravity to the particle distance to obtain a particle ratio;

[0020] Comparing the particle ratio with a first preset threshold and a second preset threshold to obtain a correction coefficient;

[0021] Adjusting the parameters in the initial flow gravity model by the correction coefficient to obtain the target flow gravity model.

[0022] In some embodiments, the method for detecting the terminal gravity values for abnormalities to obtain target abnormal gravity values comprises the following steps:

[0023] Sorting the terminal gravity values from small to large;

[0024] Screening the sorted terminal gravity values by an abnormality judgment model to obtain preliminary abnormal gravity values;

[0025] Plotting the preliminary abnormal gravity values as an abnormal gravity value distribution graph;

[0026] Analyzing and screening the distribution of the preliminary abnormal gravity values in the abnormal gravity value distribution graph to obtain the target abnormal gravity values.

[0027] In some embodiments, the screening of the ranked terminal attraction values by the anomaly judgment model to obtain preliminary abnormal attraction values comprises the following steps:

[0028] The terminal attraction values lower than the preset lower threshold are summarized to obtain first abnormal attraction values;

[0029] The terminal attraction values higher than the preset upper threshold are summarized to obtain second abnormal attraction values;

[0030] The terminal attraction values with a change amplitude exceeding a preset change rate threshold in a previous period are summarized to obtain third abnormal attraction values;

[0031] The terminal attraction values exceeding an average value of surrounding terminal attraction values and exceeding a preset difference threshold are summarized to obtain fourth abnormal attraction values;

[0032] The first abnormal attraction values, the second abnormal attraction values, the third abnormal attraction values, and the fourth abnormal attraction values are summarized to obtain the preliminary abnormal attraction values.

[0033] In some embodiments, the monthly flowback data and the geographic auxiliary data are predicted by a hybrid prediction model to obtain a predicted flow amount, comprising the following steps:

[0034] The monthly flowback data and the geographic auxiliary data are preprocessed;

[0035] The preprocessed monthly flowback data and the geographic auxiliary data are captured by a time series model to obtain first prediction data;

[0036] The preprocessed monthly flowback data, the geographic auxiliary data, and the first prediction data are integrated by a machine learning model to obtain second prediction data;

[0037] The first prediction data and the second prediction data are weighted and averaged to obtain a predicted flow amount.

[0038] In some embodiments, the target abnormal attraction values and the predicted flow amount are subjected to abnormal flow analysis to obtain an abnormal flow area range, comprising the following steps:

[0039] The predicted flow amount is divided into risk levels;

[0040] The target abnormal attraction values are marked on a regional distribution map;

[0041] The predicted flow amount after risk level division is first distinguished on the regional distribution map to obtain a predicted flow amount risk area;

[0042] Secondly, the target abnormal gravity value is obtained by the first module, and the target abnormal gravity value is input into the mixed prediction model to obtain a predicted flow amount.

[0043] In some embodiments, the target abnormal gravity value is further processed to obtain abnormal processing data, including the following steps:

[0044] The target abnormal gravity value is corrected by a time series analysis method or multi-source data fusion correction to obtain abnormal processing data.

[0045] In some embodiments, the system further comprises:

[0046] The historical production data of the target tobacco terminal is obtained.

[0047] The historical production data is analyzed by a time series prediction model to obtain a predicted total production.

[0048] The predicted total production is calculated by a gravity coefficient formula to obtain a gravity coefficient.

[0049] The target flow gravity model is adjusted by the gravity coefficient to obtain an abnormal flow gravity prediction model.

[0050] The sales data and the production data are calculated by the abnormal flow gravity prediction model to obtain a predicted gravity value.

[0051] To achieve the above-mentioned purpose, another aspect of the embodiment of the present application proposes an analysis system based on a tobacco flow gravity model, which comprises:

[0052] The first module is used to obtain sales data and production data of a target tobacco terminal.

[0053] The second module is used to input the sales data and the production data into a target flow gravity model to obtain a terminal gravity value.

[0054] The third module is used to detect the terminal gravity value to obtain a target abnormal gravity value.

[0055] The fourth module is used to predict the monthly flow data and the geographic auxiliary data by a mixed prediction model to obtain a predicted flow amount.

[0056] The fifth module is used to perform abnormal flow analysis on the target abnormal gravity value and the predicted flow amount to obtain an abnormal flow area range.

[0057] The embodiments of the present application at least have the following beneficial effects: the present application provides an analysis method and system based on a tobacco flow gravity model, the scheme processes sales data and production data through a target flow gravity model to obtain terminal gravity values, and obtains target abnormal terminal gravity values after performing abnormal detection on the terminal gravity values, and through analysis on the terminal gravity values and the target abnormal gravity values, market trends can be more comprehensively and accurately described. This method can not only identify obvious abnormal flows, but also capture some hidden and complex illegal circulation patterns. In the process of processing the sales data and the production data, the model parameters are constantly adjusted according to the actual situation, the accuracy of detection is improved, the mutual influence between terminals and the spatial relationship of the market are fully considered, and the possible abnormal flows can be more accurately identified. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 is a flowchart of an analysis method based on a tobacco flow gravity model provided by the embodiments of the present application;

[0059] Figure 2 is a flowchart of a construction method of a target flow gravity model provided by the embodiments of the present application;

[0060] Figure 3 is a flowchart of step S300 in Figure 1

[0061] Figure 4 is a schematic diagram of an analysis system based on a tobacco flow gravity model. DETAILED DESCRIPTION

[0062] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementation described in the following exemplary embodiments does not represent all the implementations consistent with the embodiments of the present application, but is only an example of devices and methods consistent with some aspects of the embodiments of the present application.

[0063] It can be understood that the terms "first", "second", and the like used in the present application can be used herein to describe various concepts, but unless specifically stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information can also be referred to as the second information, and similarly, the second information can also be referred to as the first information. Depending on the context, the word "if" as used herein can be interpreted as "when" or "when" or "in response to determining".​

[0064] As used herein, the terms "at least one", "multiple", "each", "any", and the like, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding plurality, and any refers to any one of the plurality.

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terms used herein are merely for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0066] Before the embodiments of the present application are described in detail, first, some nouns and terms involved in the embodiments of the present application are described, and the nouns and terms involved in the embodiments of the present application are applicable to the following explanations.

[0067] Terminal attraction value: refers to the demand coefficient of each region terminal for tobacco products, the greater the terminal attraction value, the greater the demand of the region terminal for the tobacco products.

[0068] Target abnormal attraction value: refers to the demand of the terminal attraction value that is different from the normal market for the tobacco products, the target abnormal attraction value is often the terminal attraction value of the cross-regional illegal circulation of tobacco products, and the region of cross-regional illegal circulation of tobacco products is deduced according to the target abnormal attraction value.

[0069] Predicted flow: refers to the prediction of the change of the flow of tobacco products in the future period, the risk level is divided by the change of the predicted flow value, and the risk region is judged according to the divided risk level, so as to adjust the supervision strength.

[0070] Abnormal processing data: refers to the target abnormal attraction value that cannot reflect the cross-regional illegal circulation of tobacco products in the process of screening the target abnormal attraction value through the terminal attraction value, and the more reliable target abnormal attraction value is obtained by obtaining the abnormal processing data and comprehensive analysis, so as to obtain the target abnormal attraction value that can reflect the cross-regional illegal circulation of tobacco products.

[0071] Predicted attraction value: refers to the predicted attraction value obtained by predicting according to the obtained abnormal flow attraction prediction model, the demand of the future market for the tobacco products is judged by the predicted attraction value, so as to adjust the appropriate production quantity according to the predicted attraction value.

[0072] Figure 1 is the flowchart of the analysis method based on the tobacco flow attraction model provided by the embodiments of the present application, Figure 1 The method in the above can include but is not limited to including steps S100 to S400:

[0073] Step S100: Obtain sales data, production data, monthly return data, and geographic auxiliary data of the target tobacco terminal;

[0074] Step S200: Input sales data and production data into the target flow gravity model to obtain the terminal gravity value;

[0075] Step S300: Perform anomaly detection on the terminal gravity value to obtain the target anomalous gravity value;

[0076] Step S400: The monthly return flow data and geographic auxiliary data are predicted using a hybrid prediction model to obtain the predicted flow volume;

[0077] Step S500: Perform abnormal flow analysis on the target abnormal gravity value and the predicted flow rate to obtain the range of abnormal flow region.

[0078] In some embodiments, step S100 involves acquiring sales data, production data, monthly return data, and geographic auxiliary data for the target tobacco terminals. The target tobacco terminals specifically refer to cigarette terminals in the research area, which can be a city, a province, or a larger geographical region, depending on the specific application requirements. Sales data typically includes information such as sales revenue and sales volume for each terminal at different time periods, while production data includes information such as production revenue and production volume for each terminal or production unit in the corresponding time period. Monthly return data refers to data on tobacco products that should have been sold within a specific sales area but were returned to other areas in violation of regulations within a month, including information on the quantity, brand, and destination of the returned products. Geographic auxiliary data refers to the geographical distance from each region to the target tobacco terminal region, the regulatory path coefficient, and holiday data. The regulatory path coefficient is a quantitative indicator of the impact of regulatory measures on target variables (such as the degree of tobacco market standardization, the incidence of violations, etc.) along a specific regulatory path. Monthly return data and geographic auxiliary data include regional location, flow volume (in ten thousand cigarettes), distance (km), regulatory coefficient, and holidays.

[0079] In some embodiments, in step S200, the construction of the target flow gravity model, such as Figure 2 As shown, Figure 2 This is a flowchart of the method for constructing the target flow gravity model provided in the embodiments of this application. Figure 2 The method may include, but is not limited to, steps S610 to S630:

[0080] Step S610: Obtain the historical total sales and historical total production of the area to be tested;

[0081] Step S620: Process the historical total sales and historical total production using the initial flow gravity model to obtain the terminal gravity value;

[0082] Step S630: correcting the initial flow gravity model by the particle model to obtain a target flow gravity model;

[0083] In some embodiments, in step S610, historical sales data and historical production data are acquired, the total sales amount of each period is calculated based on the acquired historical sales data, and the total production amount of each period is calculated based on the historical production data. The purpose of this step is to integrate the original scattered data into summary data that can be directly used for model construction. Preferably, the calculation of the total sales amount and the total production amount can be obtained by calculating the historical sales data and the historical production data in a weighted average manner to better reflect the importance of different periods. For example, more recent data can be given a higher weight to make the model closer to the current market situation. Based on the calculated total sales amount and total production amount, a target flow gravity model is constructed. The core idea of this model is to regard each cigarette terminal as a point with a terminal gravity value, and the size of the terminal gravity value is affected by the sales amount and the production amount of the terminal. By analyzing the interaction between these points, abnormal flow conditions that may exist can be identified.

[0084] In some embodiments, in step S620, the initial flow gravity model includes a first formula, a second formula, and a third formula. The total sales amount and the total production amount are calculated by the first formula to obtain a terminal non-smoke cigarette gravity value; the terminal non-smoke cigarette gravity value is calculated by the second formula to obtain a terminal gravity; and the terminal gravity is calculated by the third formula to obtain a terminal gravity value.

[0085] Specifically, the terminal non-smoke cigarette gravity value is determined by the first formula, wherein the first formula is:

[0086]

[0087] wherein F n represents the terminal non-smoke cigarette gravity value of terminal n, C k represents the total sales amount of terminal n in the kth period, P nk represents the total production amount of terminal n in the kth period, and a represents a non-smoke cigarette gravity value influence coefficient.

[0088] The first formula considers the relationship between sales and production, and by introducing the influence coefficient a, the influence degree of sales and production on the terminal gravity value can be flexibly adjusted. Generally, the value range of a is between 0.5 and 1.5, and the specific value can be adjusted according to the actual situation. For example, when a is close to 1, the influence of sales and production on the terminal gravity value is equivalent; when a is greater than 1, the influence of production on the terminal gravity value is greater; and when a is less than 1, the influence of sales on the terminal gravity value is greater.

[0089] The attraction of the terminal is determined by a second formula, where the second formula is:

[0090]

[0091] where G N represents the attraction of terminal n, r n represents the attraction radius of terminal n, F n represents the terminal non-cigarette attraction value of terminal n in the first formula.

[0092] The second formula draws on the law of gravity in physics, linking the attraction of a terminal to the distance. The attraction radius r n can be understood as the range of influence of the terminal, which can usually be determined according to factors such as the size and location of the terminal. For example, for a large retail terminal, its attraction radius may be within a range of 5-10 kilometers; while for a small terminal, its attraction radius may only be 1-2 kilometers.

[0093] After calculating the attraction of each terminal, the third formula is used to calculate the attraction value of each terminal to other terminals, taking into account the mutual influence between terminals, which helps to more comprehensively analyze the non-cigarette abnormal flow situation. The third formula is:

[0094] L n =∑(G i );

[0095] where L n represents the terminal attraction value of terminal n received by each terminal, i.e., the terminal attraction value of terminal n in the study area, G i represents the attraction of i terminals in the study area. The third formula aggregates the influence of all terminals on a particular terminal to obtain the terminal attraction value of the terminal in the entire study area.

[0096] In some embodiments, in step S630, the initial flow attraction model is modified by a point mass model to obtain a target flow attraction model. This includes: constructing a plurality of unit mass point masses centered on the target tobacco terminal, the point masses being distributed between the target tobacco terminal and other terminals; obtaining the point mass distance of two adjacent point masses; obtaining the point mass that is farther from the target tobacco terminal being attracted by the point mass that is closer to the target tobacco terminal; calculating the ratio of the point mass attraction to the point mass distance to obtain a point mass ratio; comparing the point mass ratio with a first preset threshold and a second preset threshold to obtain a correction coefficient, adjusting the parameters in the first formula by the correction coefficient to obtain the target flow attraction model.

[0097] Specifically, a plurality of unit mass particles are constructed with the terminal in the research object as the center, and these particles are distributed between the terminal and other terminals of the research object. This processing method simplifies the complex terminal network into a particle system that can be mathematically processed. Then, in the two adjacent particles, the particle farther from the terminal of the research object receives the gravitational force F generated by the particle closer to the terminal of the research object, and the distance r between the two particles is obtained. The ratio of the gravitational force F to the distance r is calculated, that is The particle ratio is obtained. The particle ratio reflects the strength of the interaction between the particles and is an important basis for correction.

[0098] The particle ratio obtained by calculation is compared with the first preset threshold and the second preset threshold to obtain a correction coefficient. The correction coefficient is positively correlated with the non-cigarette cigarette attraction value influence coefficient of the first formula in the target flow attraction model. By obtaining the adjusted correction coefficient to adjust the non-cigarette cigarette attraction value influence coefficient, the initial flow attraction model is corrected to obtain the target flow attraction model. Specifically:

[0099] When , β' = β × (1 + δ1);

[0100] When , β' = β;

[0101] When , β' = β × (1 - δ2);

[0102] Wherein, T1 represents the first preset threshold, T2 represents the second preset threshold, β represents the initial correction coefficient, β' represents the adjusted correction coefficient, and δ1 and δ2 represent adjustment parameters.

[0103] This segmented adjustment method can more flexibly cope with correction needs in different situations. Usually, T1 can be set to 1.2, T2 can be set to 0.8, and δ1 and δ2 can be set to 0.1. The specific values of these parameters can be fine-tuned according to actual application scenarios and historical data analysis results to obtain the best correction effect. For example, assuming that the initial correction coefficient β of a terminal is 1.0, the calculated The value is 1.3. Since 1.3 > T1 (assuming T1 = 1.2), β' = β x (1 + δ1), and the adjusted β' is 1.0 * (1 + 0.1) = 1.1. This means that the influence of this terminal is considered stronger than the initial estimate, so its correction coefficient needs to be appropriately increased. This method ensures the flexibility and accuracy of the model, making the target flow gravity model more realistic. Dynamic adjustment of the correction model parameters according to actual conditions makes the target flow gravity model more accurately reflect the complex situation in the real world, enabling the model to continuously optimize itself over time and improve its ability to identify and predict abnormal flow gravity.

[0104] The target flow gravity model provides a new and effective tool for identifying and preventing illegal tobacco circulation by innovatively combining the concept of gravity in physics and the special nature of the tobacco industry. This method not only considers the two key factors of sales and production, but also introduces a flexible correction method, making the model better adapt to the complex and changing market environment. Through this method, the model built by the tobacco enterprise can more accurately identify potential illegal circulation channels, so as to take corresponding prevention and control measures to effectively maintain market order and enterprise interests.

[0105] In step S200, the sales data and production data are input into the constructed target flow gravity model to obtain the terminal gravity value.

[0106] In some embodiments, as shown in Figure 3 , Figure 3 is Figure 1 the flowchart of step S300 in Figure 1 Step S300 in

[0107] Step S310: Sort the terminal gravity values from small to large;

[0108] Step S320: Filter the sorted terminal gravity values through an anomaly judgment model to obtain preliminary abnormal gravity values;

[0109] Step S330: Draw the preliminary abnormal gravity values as an abnormal gravity value distribution graph;

[0110] Step S340: Analyze and filter the preliminary abnormal gravity values according to their distribution in the abnormal gravity value distribution graph to obtain target abnormal gravity values.

[0111] In some embodiments, in step S310, the calculated terminal gravity values are sorted in order from small to large. This sorting method helps to intuitively present the relative position and influence of each terminal in the entire study area. The sorting result can help management personnel quickly identify those terminals that may need special attention.

[0112] In some embodiments, in step S320, the sorted terminal attraction values are screened by the anomaly judgment model to obtain preliminary anomaly attraction values, including: collecting terminal attraction values lower than a preset lower threshold to obtain first anomaly attraction values; collecting terminal attraction values higher than a preset upper threshold to obtain second anomaly attraction values; collecting terminal attraction values with a change amplitude exceeding a preset change rate threshold compared to the previous period to obtain third anomaly attraction values; collecting terminal attraction values exceeding the average value of surrounding terminal attraction values and exceeding a preset difference threshold to obtain fourth anomaly attraction values; and collecting the first anomaly attraction values, the second anomaly attraction values, the third anomaly attraction values, and the fourth anomaly attraction values to obtain the preliminary anomaly attraction values.

[0113] Specifically, the anomaly judgment model includes: terminal attraction values lower than a preset lower threshold, terminal attraction values higher than a preset upper threshold, terminal attraction values with a change amplitude exceeding a preset change rate threshold compared to the previous period, and terminal attraction values with a difference exceeding a preset difference threshold compared to the average terminal attraction values of surrounding terminals.

[0114] Taking terminal attraction values lower than a preset lower threshold as an example, the preset lower threshold is set to 50% of the average terminal attraction value in the region. This means that if a certain terminal attraction value is lower than half of the average level in the region, there may be an abnormal situation that needs further investigation. This may mean that the sales volume of the terminal is abnormally low, or there is unreported sales.

[0115] For the case where the terminal attraction value is higher than the preset upper threshold, such as setting the upper threshold to 200% of the average terminal attraction value in the region. If a certain terminal attraction value exceeds twice the average level in the region, it may indicate that the terminal has abnormally high sales, or it may become a transit station for illegal tobacco circulation.

[0116] The judgment standard of the change amplitude of the terminal attraction value is also important, such as setting the preset change rate threshold to 30%. That is, if a certain terminal attraction value changes by more than 30% compared to the previous period, whether it increases or decreases, it may indicate an abnormal situation. Such sudden changes may be due to dramatic changes in market environment, or may be due to illegal activities.

[0117] The difference between the terminal attraction value and the average terminal attraction value of surrounding terminals is also an important judgment standard. For example, set the preset difference threshold to 50%, if a certain terminal attraction value differs from the average terminal attraction value of its surrounding terminals by more than 50%, it may indicate that the operating conditions or market environment of the terminal differ significantly from the surrounding, and further investigation is needed.

[0118] The preliminary abnormal gravity values are obtained by aggregating the abnormal gravity values. The specific judgment criteria and threshold settings make the abnormal detection process of the present application more objective and operable. It can be understood that the thresholds are not fixed, but can be adjusted according to the specific application scenarios and historical data analysis results to obtain the best abnormal detection effect.

[0119] In some embodiments, in steps S330 to S340, in order to more intuitively show the abnormal situation, an abnormal gravity value distribution diagram is drawn based on the screened preliminary abnormal gravity values. According to the abnormal gravity value distribution diagram, it is further judged whether there is an abnormal value. This step is equivalent to a secondary confirmation of the previous abnormal detection result, which helps to reduce the possibility of misjudgment and improve the accuracy of abnormal detection. This visual method can help decision makers better understand the spatial distribution and concentration of abnormal situations, so as to formulate more targeted control measures.

[0120] In some embodiments, after the terminal gravity value is subjected to abnormal detection to obtain a target abnormal gravity value, the method further includes performing abnormal processing on the obtained target abnormal gravity value to obtain abnormal processing data, wherein the data collection frequency is resampled according to the frequency of the target abnormal gravity value, or the target abnormal gravity value is corrected by a time series analysis method or multi-source data fusion correction to obtain the abnormal processing data.

[0121] Specifically, resampling includes increasing or decreasing the data collection frequency. For example, if it is found that a certain terminal frequently appears abnormal values, the data collection frequency of the terminal can be increased to closely monitor its changes. On the contrary, if the data of a certain area is always stable, the sampling frequency can be appropriately reduced to save resources. Data correction includes trend extrapolation or multi-source data fusion correction. For example, when performing trend extrapolation, time series analysis methods such as exponential smoothing or ARIMA model can be used to predict reasonable terminal gravity values based on historical data, and the target abnormal gravity value is replaced by the reasonable terminal gravity value. Multi-source data fusion correction uses other related data sources such as sales data of surrounding areas, market research reports, etc., to comprehensively analyze more reliable data.

[0122] If resampling is selected, after adjusting the collection frequency after resampling is completed, the sales data and production data of the target tobacco terminal are reacquired, the sales data and production data are input into the target flow gravity model, the terminal gravity value is obtained, and the terminal gravity value is subjected to anomaly detection. This cycle process will continue until the preset data quality requirement is met. If data correction is selected, the sales data and production data of the target tobacco terminal are acquired according to the obtained anomaly processing data, the sales data and production data are input into the target flow gravity model, the terminal gravity value is obtained, and the terminal gravity value is subjected to anomaly detection. This cycle process will continue until the preset data quality requirement is met. This iterative optimization method can continuously improve the accuracy and reliability of the model. The data quality requirements include the following aspects: the proportion of abnormal values is not more than 5% of the total samples, the target flow gravity model processing error is within 10%, and the target flow gravity model parameter variation amplitude is less than 1% after three consecutive iterations. These specific indicators can be adjusted according to actual application requirements to balance model precision and calculation efficiency.

[0123] In some embodiments, in step S400, the monthly flowback data and the geographic auxiliary data are predicted by a mixed prediction model to obtain a predicted flow amount. This includes: pre-processing the monthly flowback data and the geographic auxiliary data; capturing data of the pre-processed monthly flowback data and the geographic auxiliary data by a time series model to obtain first prediction data; integrating multivariate features of the pre-processed monthly flowback data, the geographic auxiliary data, and the first prediction data by a machine learning model to obtain second prediction data; and performing weighted average on the first prediction data and the second prediction data to obtain the predicted flow amount.

[0124] Specifically, the pre-processing includes missing value processing, anomaly correction, feature engineering, and data standardization. More specifically, the missing value processing includes filling the null values with 0 to assume no flow or no record; the anomaly correction needs to be combined with the background of regulatory policies, such as 160,000 cigarettes in November 2024, which is marked as a special event; the feature engineering includes time features, geographic features, regulatory path features, lag features, and moving averages, wherein the time features include months, quarters, and whether it is a holiday, the geographic features are straight-line distances to the target tobacco terminal area calculated based on geopy, the regulatory path features are regulatory coefficients assigned according to the flow path, the lag features include generating flowback values lagged by 1-3 months, and the moving averages include calculating a 6-month moving average to capture long-term trends; the data standardization includes Min-Max standardization of numerical features to eliminate dimension effects.

[0125] For the mixed prediction model, including time series model and machine learning model, wherein the time series model is used to capture the seasonality, trend and holiday effect of the data, and the machine learning model is used to integrate multivariate features and handle nonlinear relationships. Specifically, the single-city time series data is input into the time series model, in the format of date and flow volume, as well as holiday labels (such as Spring Festival, Mid-Autumn Festival). For the configuration of the time series model, Prophet is selected, annual periodicity analysis is turned on, weekly periodicity is ignored, special date influence is considered with the help of custom holiday data, and df_city data is used for fitting to mine long-term trends and holiday influence patterns. The first prediction data is obtained, which includes the flow volume prediction value for the next 12 months and the seasonal decomposition results (such as trend, holiday effect). The machine learning model is input with basic features, dynamic features and the output of the time series model (first prediction data), wherein the basic features include distance, regulatory coefficient, month and holiday, the dynamic features include lag value and moving average, and the output of the time series model includes the prediction value of future flow volume and its confidence interval. For the configuration of the machine learning model, XGBRegressor is selected for regression analysis, with the number of trees set to 200, the maximum depth set to 5, the learning rate set to 0.05, the sample sampling ratio set to 0.8, the objective function set to squared error, and the training set data X_train and y_train used for fitting. Time series cross-validation is used to avoid future data leakage, and evaluation indicators include mean squared error, mean absolute error and R2 value. The second prediction data is output by the machine learning model. The first prediction data and the second prediction data are weighted and averaged to obtain the predicted flow volume, wherein the weighted average is based on the performance of the validation set.

[0126] The predicted flow volume is obtained by the mixed prediction model, and the same period comparison is calculated based on the predicted flow volume and the same period last year to obtain the same period change. According to the same period change, the risk level is divided, for example, the predicted flow volume is 398, the same period change is +12%, and the risk level is extremely high at this time. The predicted flow volume is visualized and output, and the visualization output form includes time series decomposition chart and feature importance chart, wherein the time series decomposition chart shows the trend, seasonality and holiday effect, and the feature importance chart shows the contribution of each feature to the prediction in the XGBoost model.

[0127] Among them, the risk level division includes:

[0128] Extremely high risk: predicted flow volume > historical mean + 2σ;

[0129] High risk: historical mean + σ < predicted flow volume ≤ historical mean + 2σ;

[0130] Medium risk: historical mean - σ < predicted flow volume ≤ historical mean + σ;

[0131] Low risk: predicted flow volume ≤ historical mean - σ;

[0132] wherein σ represents a ranking parameter obtained according to an empirical value.

[0133] According to multiple experiments on the hybrid prediction model, it can be known that the distance and the regulatory coefficient have the greatest impact on the predicted flow volume, followed by the holiday. Through the hybrid prediction model, a high-risk city list is generated according to the monthly updated data, so as to achieve the purpose of real-time monitoring; by adjusting the regulatory coefficient, the predicted flow volume change under different regulatory intensities is predicted, so as to perform regulatory simulation.

[0134] In some embodiments, the structure of the time series model and the machine learning model is deployed. Specifically, for the time series model, the time series model is constructed through the Prophet library. First, the Prophet class is introduced, and then the model_prophet is instantiated. In terms of structural deployment, the yearly_seasonality is turned on to capture the annual periodicity; the weekly_seasonality is set to False to ignore the weekly cycle; the holidays parameter is used to pass in the holidays_df to consider the influence of custom holidays. Finally, the fit method is called, and the df_city data is used for model training, so that the model learns the data features and rules, and the overall structure deployment is completed. For the machine learning model, the machine learning model is constructed through the XGBRegressor of the XGBoost library. When deployed, the number of trees is first determined to be 200 to control the complexity and fitting ability of the model; the maximum depth of the tree is set to 5 to avoid overfitting; the learning rate is set to 0.05 to adjust the step size of each iteration; the sample sampling ratio is 0.8 to randomly select part of the samples for training; the objective function is the square error for the regression task. Finally, the model is fitted with the training set X_train and y_train, so that the model learns the data features and rules, and the structure deployment is completed.

[0135] In some embodiments, in step S500, the target abnormal attractive force value and the predicted flow amount are subjected to abnormal flow analysis to obtain an abnormal flow area range. Specifically, the predicted flow amount is divided into risk levels; the target abnormal attractive force value is marked on the area distribution map; the predicted flow amount after risk level division is first distinguished and marked on the area distribution map to obtain a predicted flow amount risk area; and the part where the predicted flow amount risk area and the target abnormal attractive force value overlap is second distinguished and marked to obtain the abnormal flow area range. Specifically, the area distribution map can be drawn by geographic information system (GIS) software, and appropriate map projection method and scale are selected according to actual needs to clearly show the geographical features of the area. Different colors, symbols or charts are used to intuitively represent different numerical ranges when first distinguishing and marking and second distinguishing and marking.

[0136] In some embodiments, the attractive field distribution map is drawn according to the target flow attractive force model, and the terminal attractive force value is marked on the attractive field distribution map. Based on the attractive field distribution map, the state of the terminal in the research area is analyzed. The state of the terminal includes the out-of-favor state and the in-favor state, and it is determined whether each terminal belongs to the out-of-favor terminal or the in-favor terminal. The terminal whose terminal attractive force value is lower than 0 is defined as the out-of-favor terminal, and the terminal whose terminal attractive force value is higher than 0 is defined as the in-favor terminal. This classification method is simple and intuitive, and is convenient for practical operation and understanding.

[0137] After determining the state of the terminal, a corresponding advertisement placement recommendation is given. For the out-of-favor terminal, an advertisement placement recommendation from the in-favor terminal is sent to the out-of-favor terminal. The purpose of this approach is to use the influence of the in-favor terminal to drive the business development of the out-of-favor terminal. Conversely, for the in-favor terminal, an advertisement placement recommendation for the out-of-favor terminal is sent to the in-favor terminal. This strategy aims to expand the influence range of the in-favor terminal and help improve the performance of the out-of-favor terminal.

[0138] In some embodiments, when the number of in-favor terminals is lower than a preset number, a suggestion to adjust the business strategy of each terminal in the research area is triggered. This mechanism is set to ensure that the entire research area can maintain a healthy competitive situation and prevent excessive concentration or monopoly. For example, the preset number can be set to 30% of the total number of terminals. If the proportion of in-favor terminals is lower than this threshold, it may mean that the market is too concentrated or the overall environment is not healthy. In this case, some measures may need to be taken, such as increasing support for small and medium-sized terminals or adjusting pricing strategies, to promote the healthy development of the market.

[0139] In summary, the application not only effectively identifies and handles abnormal situations, but also provides specific business recommendations based on model results. This comprehensive solution from data analysis to practical application provides a powerful support tool for market management and operational decision-making in the tobacco industry. By continuously applying and optimizing the target flow gravity model, market dynamics can be better understood, illegal circulation can be effectively prevented, and more precise market strategies can be developed.

[0140] In some embodiments, further comprising: obtaining historical production amount data of the target tobacco terminal; calculating the historical production amount data through a fourth formula or a time series prediction model to obtain a predicted total production amount; calculating the predicted total production amount through a gravity coefficient formula to obtain a gravity coefficient; adjusting the target flow gravity model through the gravity coefficient to obtain an abnormal flow gravity prediction model; and calculating the sales data and the production data through the abnormal flow gravity prediction model to obtain a predicted gravity value. Through the abnormal flow gravity prediction model, not only the current market situation can be analyzed, but also the future trend can be reasonably predicted, providing an important basis for long-term strategic decision-making of the enterprise.

[0141] Specifically, historical production amount data of each terminal in each time period in the research area is obtained. These historical data are the basis for prediction, usually covering a time span of 12 to 36 months in the past, in order to capture possible seasonal fluctuations and long-term trends. Based on the historical production amount data, the predicted total production amount of each terminal is calculated through the fourth formula. The fourth formula is:

[0142]

[0143] where P n represents the predicted total production amount of terminal n, P nk represents the historical production amount data of terminal n in the kth time period, and K represents the total number of time periods. The fourth formula is suitable for relatively stable market environments. However, in actual applications, more complex prediction models such as exponential smoothing or ARIMA models can also be selected according to specific circumstances to improve the accuracy of prediction.

[0144] Based on the predicted total production amount, the terminal gravity coefficient is adjusted through the gravity coefficient formula, which is:

[0145]

[0146] where β n represents the gravity coefficient of terminal n, P′ n represents the predicted total production amount of terminal n, and P nLet γ represent the rated total production of terminal n, and let γ represent the adjustment parameter. When the predicted total production exceeds the rated total production, the gravitational coefficient of the terminal is increased; conversely, it is decreased. The value of the adjustment parameter γ is typically between 0.5 and 1.5 and can be adjusted according to actual conditions. For example, when γ = 1, the change in the gravitational coefficient is proportional to the change in total production; when γ < 1, the change in the gravitational coefficient is relatively gradual; and when γ > 1, the change in the gravitational coefficient is amplified.

[0147] Assume a certain terminal has a rated total production volume P n The production volume is 1000 units, while the total production volume P′ predicted based on historical data is... n The value is 1200 units, and the adjustment parameter γ is set to 1. Substituting the specific value into the formula for the gravitational coefficient yields... This means that the terminal's gravity coefficient is adjusted to 1.2 because the predicted total production is higher than the rated total production. This indicates that the terminal's influence is expected to increase in the future period because it is expected to produce more products, thus increasing its attractiveness in the market. Based on the abnormal flow gravity prediction model, the predicted gravity values ​​for each terminal in the future period are calculated. This dynamic adjustment method allows the model to better adapt to market changes, improving the accuracy and reliability of predictions.

[0148] In some embodiments, the method further includes: acquiring historical consumption data, historical sales data, and historical production data of the target tobacco terminal; quantifying the linear relationship of the historical consumption data, historical sales data, and historical production data using the Pearson correlation coefficient to obtain the correlation; measuring the parity of the historical consumption data, historical sales data, and historical production data by calculating the number of consecutive months of increase or decrease; measuring the dispersion of the historical consumption data, historical sales data, and historical production data by calculating the coefficient of variation; and identifying abnormal production states by analyzing the correlation, parity, and dispersion. This multi-dimensional data provides a rich source of information for analysis. Typically, historical consumption data, historical sales data, and historical production data should cover a time span of at least 12 months to capture potential seasonal variations. Historical consumption data represents the amount of money consumers spend on goods, focusing on reflecting consumers' purchasing power and demand. Historical sales data represents the revenue generated by businesses selling goods, focusing more on the business's sales performance and market performance.

[0149] Specifically, the Pearson correlation coefficient is used to quantify the linear relationship between historical consumption data, historical sales data, and historical production data. If the correlation coefficient between consumption and sales is close to 1, while the correlation between production and these two is low, this may indicate inventory backlog or sales through irregular channels.

[0150] The odd-even degree is measured by counting the number of consecutive rising or falling months, and the dispersion degree is measured by calculating the coefficient of variation (standard deviation divided by the mean). These indicators can help identify abnormal patterns in the data. For example, if the sales data of a certain terminal shows a high degree of dispersion, while the consumption and production are relatively stable, it may indicate that there are irregular large transactions in that terminal.

[0151] For abnormal production state by analyzing the correlation, odd-even degree and dispersion degree. Specifically, the abnormal production state is divided into two categories: stable production and abnormal production. Stable production is further divided into benign production and implicit abnormal production, while abnormal production is divided into obvious production anomaly and invisible production anomaly. For example, when the correlation of historical consumption data, historical sales data and historical production data is stable, and the odd-even degree and dispersion degree are within the normal range, it can be determined as benign production. If the correlation is stable but the dispersion degree is slightly high, it may belong to implicit abnormal production. When the correlation changes significantly or the dispersion degree of a certain indicator is abnormally high, it can be determined as obvious production anomaly. When the correlation and dispersion degree are both in a critical state, it may belong to invisible production anomaly. This multi-level classification method can more accurately depict the production state of the terminal, and help to discover potential problems and take appropriate management measures in time.

[0152] In some embodiments, the application of the tobacco flow gravity model-based analysis method provides a powerful support tool for market management of the tobacco industry.

[0153] Specifically, the consumption data and production data of the tobacco terminal in the research object area are obtained. These data are the basis of the entire analysis process, and can usually be obtained through the enterprise's sales system, production system and market research, etc. By inputting the sales data and production data into the target flow gravity model constructed, the terminal gravity value is obtained, the terminal gravity value is subjected to abnormal detection, the target abnormal gravity value is obtained, the terminal gravity value and the target abnormal gravity value are subjected to abnormal flow analysis, the abnormal flow trend is obtained, and the applications that can be carried out according to the abnormal flow trend include abnormal flow analysis, market strategy formulation, inventory optimization, risk assessment, resource allocation and performance evaluation.

[0154] Specifically, abnormal flow analysis: By analyzing the distribution and trend of terminal gravity values, abnormal flow patterns in the study area are identified. For example, if a terminal gravity value suddenly rises sharply while the surrounding terminal gravity values decline, it may mean that there is illegal cross-regional transfer behavior. By discovering these abnormal patterns in a timely manner, enterprises can take appropriate measures to prevent and combat illegal circulation channels. Market strategy formulation: Based on the distribution of terminal gravity values, enterprises can formulate differentiated marketing strategies. For example, for terminals with high terminal gravity values, they can consider increasing product categories or providing more promotional support; while for terminals with low terminal gravity values, they may need to take measures such as improving service quality or adjusting product structure. Inventory optimization: Based on the prediction of future changes in terminal gravity values, enterprises can more scientifically manage inventory. For example, for terminals with predicted rising terminal gravity values, inventory levels can be appropriately increased; while for terminals with predicted declining terminal gravity values, inventory can be reduced to avoid capital occupation. Risk assessment: Combined with the results of abnormal production state analysis, enterprises can comprehensively assess the operating risks of each terminal. For example, for terminals determined to have hidden production abnormalities, supervision and audit efforts may need to be strengthened to prevent potential violations. Resource allocation: Based on terminal gravity values and abnormal states, enterprises can more reasonably allocate various resources such as marketing funds and human resources. For example, more resources can be invested in terminals with high growth potential but currently low terminal gravity values to promote their rapid development. Performance evaluation: Terminal gravity value changes are used as one of the important indicators for performance evaluation. For example, the increase in terminal gravity value can be included in the KPI system of terminal managers to encourage them to take effective measures to improve the market influence of the terminal.

[0155] Through these specific applications, enterprises can comprehensively improve the efficiency and accuracy of market management. For example, a tobacco company successfully identified five terminals with abnormal flow after applying this method, timely blocking an important illegal circulation channel, resulting in a 3% increase in legal sales the following year. Another company reduced inventory turnover days from 45 to 35, significantly reducing capital occupation. In general, the analysis method based on the tobacco flow gravity model provides a comprehensive, systematic, and efficient market management tool for the tobacco industry. By considering factors such as consumption, production, and sales, combined with advanced data analysis techniques, this method can help enterprises better grasp market dynamics, effectively prevent illegal circulation, optimize resource allocation, and thus occupy a dominant position in the fierce market competition.

[0156] As shown in Figure 4 , an analysis system based on a tobacco flow gravity model includes:

[0157] The first module is configured to acquire sales data and production data of a target tobacco terminal;

[0158] The second module is configured to input the sales data and the production data into a target flow gravity model to obtain a terminal gravity value;

[0159] The third module is configured to perform anomaly detection on the terminal gravity value to obtain a target anomaly gravity value;

[0160] The fourth module is configured to perform prediction on monthly flow back data and geographic auxiliary data by using a hybrid prediction model to obtain a predicted flow amount;

[0161] The fifth module is configured to perform anomaly flow analysis on the target anomaly gravity value and the predicted flow amount to obtain an anomaly flow area range.

[0162] It can be understood that the contents in the method embodiments are applicable to the system embodiments, the system embodiments specifically implement the functions of the method embodiments, and achieve the same beneficial effects as the method embodiments.

[0163] The embodiments described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. It can be understood by those skilled in the art that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0164] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and can include more or fewer steps than the figures, or combine certain steps or different steps.

[0165] Those skilled in the art can understand that all or some steps in the method disclosed above, the functions of the modules / units in the system and the device can be implemented as software, firmware, hardware and appropriate combinations thereof.

[0166] The terms "first", "second", "third", "fourth" and the like (if any) in the specification of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0167] It should be understood that in the present application, "at least one" refers to one or more, and "multiple" refers to two or more. "And / or" is used to describe the association relationship of the associated objects, which means that there can be three relationships, for example, "A and / or B" can represent three cases of only A, only B, and A and B existing at the same time, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after it. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can represent a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0168] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0169] If the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program storage media.

[0170] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, but this does not limit the scope of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. An analytical method based on a tobacco flow gravity model, characterized in that, The method includes: Acquire sales data, production data, monthly return data, and geographic auxiliary data of the target tobacco terminals; The sales data and production data are input into the target flow gravity model to obtain the terminal gravity value; Anomaly detection is performed on the terminal gravity value to obtain the target abnormal gravity value; The predicted flow volume is obtained by using a hybrid prediction model to predict the monthly return flow data and the geographic auxiliary data. The predicted flow is classified into risk levels; the target anomalous gravity value is marked on a regional distribution map; the predicted flow after risk level classification is marked with a first distinction on the regional distribution map to obtain the predicted flow risk area; the part where the predicted flow risk area and the target anomalous gravity value overlap is marked with a second distinction to obtain the range of the anomalous flow area; The monthly return data refers to the quantity, brand, and flow direction of tobacco products that should have been sold in a specific sales area within a month but were illegally returned to other areas; the geographical auxiliary data refers to the geographical distance from each region to the target tobacco terminal, the regulatory path coefficient, holiday marking data, and regional location information; the terminal attraction value represents the demand coefficient of each region's terminals for tobacco products, and the larger the terminal attraction value, the greater the demand for the tobacco product in that region's terminals; the target abnormal attraction value represents the demand for the tobacco product that differs from the normal market demand in the terminal attraction value.

2. The method according to claim 1, characterized in that, The method for constructing the target flow gravity model includes the following steps: Obtain the historical total sales and historical total production of the area to be tested; The historical sales volume and the historical production volume are processed using an initial flow gravity model to obtain the terminal gravity value; The initial flow gravity model is modified using a point mass model to obtain the target flow gravity model.

3. The method according to claim 2, characterized in that, The process of modifying the initial flow gravity model using a point mass model to obtain the target flow gravity model includes the following steps: Multiple mass points of unit mass are constructed with the target tobacco terminal as the center, and the mass points are distributed between the target tobacco terminal and other terminals; Obtain the particle distance between two adjacent particles; In obtaining the particle distance, the particle farther from the target tobacco terminal is subject to the gravitational force of the particle closer to the target tobacco terminal; Calculate the ratio of the gravitational force of the particle to the distance to the particle to obtain the particle ratio; The correction coefficient is obtained by comparing the particle ratio with a first preset threshold and a second preset threshold. The target flow gravity model is obtained by adjusting the parameters in the initial flow gravity model using the correction coefficient.

4. The method according to claim 1, characterized in that, The process of detecting anomalies in the terminal's gravitational value to obtain a target anomalous gravitational value includes the following steps: Sort the terminal gravity values ​​from smallest to largest; The sorted terminal gravity values ​​are filtered using an anomaly detection model to obtain preliminary anomaly gravity values; The preliminary abnormal gravity values ​​were plotted as an abnormal gravity value distribution map; The target anomalous gravity value is obtained by analyzing and filtering the distribution of the preliminary anomalous gravity values ​​in the anomalous gravity value distribution map.

5. The method according to claim 4, characterized in that, The step of filtering the sorted terminal gravity values ​​using an anomaly detection model to obtain preliminary anomaly gravity values ​​includes the following steps: The terminal gravity values ​​that are lower than a preset lower threshold are aggregated to obtain the first abnormal gravity value; The terminal gravity values ​​that exceed the preset upper limit threshold are aggregated to obtain the second abnormal gravity value; The terminal gravity values ​​that change more than a preset rate of change threshold compared to the previous period are summarized to obtain the third abnormal gravity value. The terminal gravity values ​​that exceed the average value of the surrounding terminal gravity values ​​and exceed a preset difference threshold are aggregated to obtain the fourth abnormal gravity value. The first, second, third, and fourth anomalous gravitational values ​​are summarized to obtain the preliminary anomalous gravitational value.

6. The method according to claim 1, characterized in that, The predicted flow is obtained by using a hybrid prediction model to predict the monthly return flow data and the geographic auxiliary data, including the following steps: The monthly return data and the geographic auxiliary data are preprocessed; The first prediction data is obtained by capturing data from the preprocessed monthly return data and the geographic auxiliary data using a time series model. The second prediction data is obtained by integrating the preprocessed monthly return data, the geographic auxiliary data, and the first prediction data using a machine learning model with multivariate features. The predicted flow rate is obtained by taking a weighted average of the first and second predicted data.

7. The method according to claim 1, characterized in that, It also includes anomaly processing of the obtained target anomalous gravity value to obtain anomaly processing data, including the following steps: The target anomalous gravity value is corrected by time series analysis or multi-source data fusion correction to obtain anomalous processing data.

8. The method according to claim 1, characterized in that, Also includes: Obtain the historical production data of the target tobacco terminal; The historical production data is analyzed using a time series forecasting model to obtain the predicted total production. The gravity coefficient is obtained by calculating the predicted total production using the gravity coefficient formula. By adjusting the target flow gravity model using the gravity coefficient, an abnormal flow gravity prediction model is obtained. The sales data and production data are calculated using the abnormal flow gravity prediction model to obtain a predicted gravity value. The predicted gravity value represents the predicted gravity value obtained by predicting based on the acquired abnormal flow gravity prediction model. The predicted gravity value is used to determine the future market demand for the tobacco product, and the appropriate production quantity is adjusted accordingly.

9. An analysis system based on a tobacco flow gravity model, characterized in that, The system includes: The first module is used to acquire sales data, production data, monthly return data, and geographic auxiliary data of the target tobacco terminal; wherein the monthly return data represents the quantity, brand, and flow direction of tobacco products that should have been sold in a specific sales area within a month but were illegally returned to other areas; the geographic auxiliary data represents the geographic distance from each area to the target tobacco terminal, the regulatory path coefficient, holiday marking data, and regional location information. The second module is used to input the sales data and the production data into the target flow gravity model to obtain the terminal gravity value; wherein the terminal gravity value represents the demand coefficient of tobacco products for terminals in each region. The larger the terminal gravity value, the greater the demand for tobacco products for terminals in that region. The third module is used to detect anomalies in the terminal gravity value and obtain a target abnormal gravity value; wherein the target abnormal gravity value represents the demand for the tobacco product in the terminal gravity value that is different from that in the normal market. The fourth module is used to predict the monthly return flow data and the geographic auxiliary data using a hybrid prediction model to obtain the predicted flow volume; The fifth module is used to classify the predicted flow into risk levels; mark the target anomalous gravity value on the regional distribution map; perform a first differentiation mark on the regional distribution map after classifying the risk levels to obtain the predicted flow risk area; and perform a second differentiation mark on the part where the predicted flow risk area and the target anomalous gravity value overlap to obtain the range of the anomalous flow area.

Images