A Machine Learning-Based Data Value Analysis Method and System

Abstract

The present invention relates to the technical field of machine learning, and discloses a data value analysis method and system based on machine learning. The method includes: collecting customer data of an agricultural product e-commerce platform; generating a keyword group of the customer according to the customer data; generating a purchase intention combination of the customer based on a decision tree model and the keyword group; performing importance analysis on the keyword group according to the purchase intention combination to obtain an importance score of the keywords in the keyword group; screening out target words in the keywords according to a screening mechanism of the importance score; and generating a predicted sales volume of agricultural products according to the target words and the customer data. The present invention can optimize market strategies based on machine learning and maximize data value.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and particularly to a method and system for analyzing data value based on machine learning. Background Art

[0002] In the agricultural product e-commerce platform and supply chain management, sales forecasting is an important basis for determining inventory management, production scheduling, and marketing strategies. Existing technologies usually adopt static sales forecasting models based on historical data, mainly including time series analysis, regression analysis, and prediction methods based on machine learning. These methods rely on fixed historical data sets to establish static models to predict future sales volume.

[0003] In the process of data-driven value mining, the dynamic evolution characteristics of market demand play a decisive role in the prediction accuracy. However, there are significant bottlenecks in the traditional static prediction mode in fully releasing data value, which are mainly reflected in the following two aspects:

[0004] Firstly, external interference factors such as promotional activities have a strong influence on market demand. However, due to the lack of a real-time response mechanism to such dynamic changes in the static prediction method, it is impossible to timely obtain and update the value labels of agricultural products, resulting in a lag effect between the prediction results and the actual market demand, and it is difficult to fit the immediate state of the market. Secondly, existing methods usually use a fixed-length time window for prediction, and the historical data sampling of the fixed time window has limitations. When the market demand fluctuates greatly, the short-term market changes are difficult to be captured by the fixed window in time, resulting in a slow response of the prediction model to the dynamic changes of the market. For example, if the window length is too long, the influence of the latest market trend is diluted, and if the window is too short, the model is easily affected by short-term fluctuations, resulting in unstable predictions. Summary of the Invention

[0005] The present invention provides a method and system for analyzing data value based on machine learning, and its main purpose is to solve the problem in the prior art that due to the lack of a real-time response mechanism to dynamic changes, it is impossible to timely obtain and update the value labels of agricultural products, resulting in a lag effect between the prediction results and the actual market demand.

[0006] To achieve the above object, a method for analyzing data value based on machine learning provided by the present invention includes:

[0007] Collect customer data of the agricultural product e-commerce platform;

[0008] Generate keyword groups of customers according to the customer data;

[0009] Generate the purchase intention combination of the customers based on the decision tree model and the keyword groups;

[0010] Perform importance analysis on the keyword group according to the purchase intention combination to obtain the importance scores of the keywords in the keyword group;

[0011] Screen out the target words among the keywords according to the screening mechanism of the importance scores;

[0012] Generate the predicted sales volume of agricultural products according to the target words and the customer data;

[0013] Monitor the real-time sales volume of the agricultural products, update the importance scores of the target words based on the predicted sales volume and the real-time sales volume, and update the target words according to the updated importance scores to obtain the value labels of the agricultural products.

[0014] Optionally, the generating the purchase intention combination of the customer based on the decision tree model and the keyword group includes:

[0015] T1. Generate the data set of the customer according to the keywords and the customer data, where the keywords include: agricultural product A, agricultural product B, and agricultural product C, and the customer data includes: purchase history, browsing time, and browsing times of agricultural products;

[0016] T2. Input the data set into the decision tree model;

[0017] T3. When the keyword has a purchase history, mark the keyword as Y. When the keyword has no purchase history, enter the browsing time judgment algorithm, where the browsing time judgment algorithm is:

[0018] When the browsing time of the keyword ≥ 10, mark the keyword as Y;

[0019] When the browsing time of the keyword is within 2 - 10s, enter the browsing times judgment algorithm, where the browsing times judgment algorithm is: when the browsing times of the keyword ≥ 3 times, mark the keyword as Y, and when the browsing times of the keyword < 3 times, mark the keyword as N;

[0020] When the browsing time of the keyword ≤ 2s, mark the keyword as N;

[0021] T4. Add the keywords marked as Y to the purchase intention combination of the customer.

[0022] Optionally, the performing importance analysis on the keyword group according to the purchase intention combination to obtain the importance scores of the keywords in the keyword group includes:

[0023] Generate the purchase history impact score, browsing time impact score, and browsing frequency impact score for the keywords in the keyword group according to the purchase intention combination, where:

[0024] The calculation formula for the purchase history impact score is as follows:

[0025]

[0026] Where, represents the th keyword in the keyword group, () represents the purchase history impact score, represents the customer, represents the purchase history of the keyword, represents that the customer has a purchase history of the keyword, represents that the customer does not have a purchase history of the keyword;

[0027] The calculation formula for the browsing time impact score is as follows:

[0028]

[0029] Where, represents the th keyword in the keyword group, represents the browsing time impact score, represents the browsing time of the keyword;

[0030] The calculation formula for the browsing frequency impact score is as follows:

[0031]

[0032] Where, represents the th keyword in the keyword group, represents the browsing frequency impact score, represents the browsing frequency of the keyword;

[0033] Perform a weighted sum of the purchase history impact score, the browsing time impact score, and the browsing frequency impact score to obtain the importance score of the keyword, where the calculation formula for the weighted sum is as follows:

[0034]

[0035] Where, represents the th keyword in the keyword group, represents the importance score of the keyword, 、 , respectively represent the weight coefficients of the purchase history impact score, the browsing time impact score, and the browsing frequency impact score.

[0036] Optionally, the screening mechanism is as follows:

[0037]

[0038] where represents the th keyword in the keyword group, represents a preset importance threshold, represents the target word after screening, represents the importance score.

[0039] Optionally, generating the predicted sales volume of agricultural products based on the target word and the customer data includes:

[0040] Substituting the target word and the customer data into the predicted sales volume algorithm to obtain the predicted sales volume of agricultural products, where the predicted sales volume algorithm is as follows:

[0041]

[0042] where represents the th keyword in the keyword group, represents the word vector corresponding to the target word, represents the data vector corresponding to the customer data, represents the predicted sales volume corresponding to the th keyword, represents the mapping function.

[0043] Optionally, monitoring the real-time sales volume of the agricultural products and updating the importance score of the target word based on the predicted sales volume and the real-time sales volume includes:

[0044] Collecting the sales volume data of the agricultural products in the current month based on the sliding window algorithm to obtain the real-time sales volume of the agricultural products in the current month;

[0045] Updating the importance score of the target word based on the predicted sales volume and the real-time sales volume, where the update formula of the importance score is as follows:

[0046]

[0047] where represents the th keyword in the keyword group, Represents the importance score of the target word, Represents the real-time sales volume, Represents the predicted sales volume, Represents the total number of days in the current month, Represents the total number of days for collecting the sales volume data, Represents the promotion indicator variable, Represents the correction coefficient of the real-time sales volume error, Represents the correction coefficient of the impact of promotional activities, Represents the updated importance score.

[0048] Optionally, the calculation formula of the sliding window algorithm is as follows:

[0049]

[0050] Wherein, Represents the total number of days in the current month, Represents the total number of days for collecting the sales volume data, Represents the adjustment coefficient, Represents rounding down, Represents the minimum number of days for data collection, Represents the maximum number of days for data collection.

[0051] Optionally, the calculation formula of the adjustment coefficient is as follows:

[0052]

[0053] Wherein, Represents the adjustment coefficient, Represents a random number.

[0054] Optionally, the calculation formula of the real-time sales volume is as follows:

[0055]

[0056] Wherein, Represents the real-time sales volume on the first day of the current month, Represents the real-time sales volume on the second day of the current month, Represents the real-time sales volume on the nth day of the current month, Represents the real-time sales volume.

[0057] To solve the above problems, the present invention also provides a data value analysis system based on machine learning, and the system includes:

[0058] A collection module for collecting customer data of an agricultural product e-commerce platform;

[0059] A keyword group generation module for generating keyword groups of customers according to the customer data;

[0060] A purchase intention combination generating module, used for generating the customer's purchase intention combination based on a decision tree model and the keyword group;

[0061] An importance score generating module, used for performing importance analysis on the keyword group according to the purchase intention combination to obtain importance scores of the keywords in the keyword group;

[0062] A screening module, used to screen out target words from the keywords according to the screening mechanism of the importance score;

[0063] A sales forecasting module, used for generating a forecasted sales volume of agricultural products according to the target words and the customer data;

[0064] The value label generation module is used to monitor the real-time sales volume of the agricultural product, update the importance score of the target word based on the predicted sales volume and the real-time sales volume, and update the target word according to the updated importance score to obtain the value label of the agricultural product.

[0065] Compared with the prior art, the present invention has the following beneficial effects:

[0066] 1. When the relationship between real-time sales and predicted sales changes, the importance score of the target word is increased or decreased accordingly, and the keywords are screened and sorted again, and the keywords corresponding to the top three importance scores are selected as value tags. This real-time monitoring and dynamic update mechanism can adjust the value tags in time according to the actual sales situation of the market, so that the value tags can reflect the actual needs of the market in real time, avoiding the lag effect between the prediction results and the actual needs of the market;

[0067] 2. Use the sliding window algorithm to collect the sales data of agricultural products in the current month to obtain real-time sales. By setting the minimum and maximum data collection days to limit the size of the sliding window, it ensures that the collected data has a sufficient time span to capture long-term trends such as seasonal changes and cyclical fluctuations, avoiding unstable or misleading predictions due to insufficient data collection, and avoiding the problem of historical data diluting the latest market dynamics due to too long a window, so that the model has a better response capability to the latest market trends;

[0068] 3. The total number of days for sales data collection can be randomly selected between the minimum and maximum data collection days, effectively avoiding the situation where the updated importance score fluctuates dramatically due to abnormal data in a fixed window. In this way, the update formula of the importance score can be more robust, successfully avoiding the interference of extreme data values ​​on the update formula of the importance score, and ensuring the stability and reliability of the update process of the importance score. Brief Description of the Drawings

[0069] Figure 1 FIG. 1 is a schematic flow chart of a method for analyzing data value based on machine learning provided by an embodiment of the present invention;

[0070] Figure 2 FIG. 2 is a functional module diagram of a data value analysis system based on machine learning provided by an embodiment of the present invention;

[0071] The implementation, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed Embodiments

[0072] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0073] An embodiment of the present application provides a method for analyzing data value based on machine learning. The execution subject of the method for analyzing data value based on machine learning includes, but is not limited to, at least one of electronic devices such as a server, a terminal, etc. that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for analyzing data value based on machine learning can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0074] Referring to Figure 1 FIG. 1, which is a schematic flow chart of a method for analyzing data value based on machine learning provided by an embodiment of the present invention. In this embodiment, the method for analyzing data value based on machine learning includes:

[0075] S1. Collect customer data of the agricultural product e-commerce platform.

[0076] In the embodiment of the present invention, the collection of customer data of the agricultural product e-commerce platform includes:

[0077] Using the background log and front-end Cookie technology of the agricultural product e-commerce platform to collect customer data of customers on the agricultural product e-commerce platform, where the customer data is the purchase history, browsing time and browsing times of agricultural products.

[0078] Specifically, the customer data of the agricultural product e-commerce platform is collected through the background logs and front-end Cookie technology of the agricultural product e-commerce platform, including the purchase history, browsing time, and browsing times of agricultural products. Among them, the purchase history of agricultural products details the customer's order information, including the types, quantities, transaction times, and payment amounts of the purchased products; the browsing time of agricultural products records the time when the customer visits the product details page; the browsing times of agricultural products record the cumulative number of times the customer browses a certain agricultural product on the agricultural product e-commerce platform. Customer data can not only help the agricultural product e-commerce platform better understand consumer needs and behavior patterns, but also be transformed into actual business value through data mining and machine learning technologies, promoting the agricultural product e-commerce platform to achieve precision marketing, scientific inventory management, and dynamic decision optimization.

[0079] For example, through the front-end Cookie technology, customer data is collected. The data record shows that the customer has a purchase history of a certain agricultural product, indicating that the customer has successfully purchased the product and has certain consumption preferences; at the same time, the data also records that the customer's page browsing time for this agricultural product is 20 seconds, indicating that the customer shows a high degree of attention during the browsing process; in addition, the customer's cumulative browsing times for this product reach 5 times, further reflecting their potential purchase intention and continuous interest.

[0080] Combining the above information, customer data not only reflects historical purchase behaviors, but also reveals the degree of attention to products through browsing time and browsing times. This information is of great significance for constructing accurate customer portraits. Based on this, the agricultural product e-commerce platform can use customer data for targeted recommendations and precision marketing, while providing strong decision-making support for sales forecasting and inventory management, further improving the operation efficiency and customer conversion rate of the agricultural product e-commerce platform.

[0081] S2. Generate keyword groups for the customer according to the customer data.

[0082] In the embodiment of the present invention, the generating keyword groups for the customer according to the customer data includes:

[0083] Remove the noise data, duplicate records, and irrelevant data in the customer data to ensure the accuracy of the customer data, and perform unified coding processing on the customer data (for example, convert to lowercase uniformly and remove special symbols) to provide a standardized input for word segmentation and subsequent analysis;

[0084] Extract features related to customer behaviors and preferences from the customer data, such as extracting the types, brands, specifications, purchase frequencies, etc. of the purchased agricultural products from the purchase history of agricultural products, and extracting the degree of attention of the customer to different agricultural product pages from the browsing time and browsing times of agricultural products;

[0085] When there is text information such as evaluations and messages in the customer data, perform word segmentation, part-of-speech tagging, stop word removal, etc. on the text information to facilitate subsequent keyword extraction;

[0086] Use a keyword extraction algorithm (such as the SF-IDF algorithm) to extract important words from the processed customer data to form a keyword group. The expression of the keyword group is as follows:

[0087]

[0088] Among them, represents the first keyword, represents the nth keyword, where n represents the number of keywords, represents the keyword group.

[0089] S3. Generate the purchase intention combination of the customer based on the decision tree model and the keyword group.

[0090] In the embodiment of the present invention, generating the purchase intention combination of the customer based on the decision tree model and the keyword group includes:

[0091] T1. Generate a data set of the customer according to the keyword and the customer data. Among them, the keyword includes: agricultural product A, agricultural product B, and agricultural product C, and the customer data includes: purchase history of agricultural products, browsing time, and browsing times;

[0092] T2. Input the data set into the decision tree model;

[0093] T3. When the keyword has a purchase history, mark the keyword as Y. When the keyword has no purchase history, enter the browsing time judgment algorithm. The browsing time judgment algorithm is as follows:

[0094] When the browsing time of the keyword ≥ 10, mark the keyword as Y;

[0095] When the browsing time of the keyword is within 2 - 10s, enter the browsing times judgment algorithm. The browsing times judgment algorithm is: when the browsing times of the keyword ≥ 3 times, mark the keyword as Y. When the browsing times of the keyword < 3 times, mark the keyword as N;

[0096] When the browsing time of the keyword ≤ 2s, mark the keyword as N;

[0097] T4. Add the keyword marked as Y to the purchase intention combination of the customer.

[0098] Specifically, a dataset of the customer is generated according to the keywords and the customer data. Wherein, taking the customer as a unit, the corresponding keywords are associated with the purchase history of the customer's agricultural products (whether the relevant agricultural products have been purchased), the browsing time (the browsing duration of the agricultural products corresponding to the keywords), and the browsing times (the browsing times of the agricultural products corresponding to each keyword), forming a set containing the keywords and the corresponding customer data, and this set is the dataset of the customer.

[0099] For example, taking customer 1 as an example, the dataset of customer 1 may include the keyword "agricultural product A" and the purchase history of customer 1 for agricultural product A (assuming purchased), the browsing time (such as 30 minutes), and the browsing times (such as 5 times).

[0100] S4. Perform importance analysis on the keyword group according to the purchase intention combination to obtain the importance scores of the keywords in the keyword group.

[0101] In the embodiment of the present invention, the performing importance analysis on the keyword group according to the purchase intention combination to obtain the importance scores of the keywords in the keyword group includes:

[0102] Generate the purchase history influence score, the browsing time influence score, and the browsing times influence score of the keywords in the keyword group according to the purchase intention combination, where:

[0103] The calculation formula of the purchase history influence score is as follows:

[0104]

[0105] Wherein, represents the th keyword in the keyword group, () represents the purchase history influence score, represents the customer, represents the purchase history of the keyword, represents that the customer has the purchase history of the keyword, represents that the customer does not have the purchase history of the keyword;

[0106] The calculation formula of the browsing time influence score is as follows:

[0107]

[0108] Wherein, represents the th keyword in the keyword group, represents the browsing time influence score, Indicates the browsing time of the keyword;

[0109] The calculation formula for the influence of the browsing times on the score is as follows:

[0110]

[0111] Wherein, Indicates the th keyword in the keyword group, Indicates the influence score of the browsing times, Indicates the browsing times of the keyword;

[0112] Perform weighted summation on the influence score of the purchase history, the influence score of the browsing time, and the influence score of the browsing times to obtain the importance score of the keyword. Wherein, the calculation formula for weighted summation is as follows:

[0113]

[0114] Wherein, Indicates the th keyword in the keyword group, Indicates the importance score of the keyword, , , respectively indicate the weight coefficients of the influence score of the purchase history, the influence score of the browsing time, and the influence score of the browsing times.

[0115] Specifically, , and 1, , , The values of can be dynamically adjusted according to the goals of the application. If a certain scoring factor (such as purchase history) is more influential than other factors (such as browsing times or time), the weighted coefficients can be adjusted to make the model more sensitive to this scoring factor.

[0116] For example, when the purchase history is very critical in predicting the purchase intention, the value of is set to be larger. When the influence of the browsing time on the purchase intention is smaller, the value of is appropriately reduced, and the browsing times are adjusted according to the results of actual analysis.

[0117] S5. Screen out the target words among the keywords according to the screening mechanism of the importance score.

[0118] In the embodiment of the present invention, the screening mechanism is as follows:

[0119]

[0120] Among them, represents the th keyword in the keyword group, represents a preset importance threshold, represents the target word after screening, represents the importance score.

[0121] S6. Generate a predicted sales volume of agricultural products based on the target word and the customer data.

[0122] In the embodiment of the present invention, generating the predicted sales volume of agricultural products based on the target word and the customer data includes:

[0123] Substitute the target word and the customer data into the predicted sales volume algorithm to obtain the predicted sales volume of agricultural products, where the predicted sales volume algorithm is as follows:

[0124]

[0125] Among them, represents the th keyword in the keyword group, represents the word vector corresponding to the target word, represents the data vector corresponding to the customer data, represents the predicted sales volume corresponding to the th keyword, represents the mapping function.

[0126] Specifically, represents the mapping function of the neural network, is a model, The specific structure and form depend on the architecture of the neural network, Learn the complex mapping relationship from input to output, process the input data (target word vector and customer data vector) through multiple layers (convolutional layer, fully connected layer, etc.), and finally output the predicted sales volume.

[0127] S7. Monitor the real-time sales volume of the agricultural products, update the importance score of the target word based on the predicted sales volume and the real-time sales volume, and update the target word according to the updated importance score to obtain the value label of the agricultural products.

[0128] In the embodiment of the present invention, monitoring the real-time sales volume of the agricultural products and updating the importance score of the target word based on the predicted sales volume and the real-time sales volume includes:

[0129] Collect data on the sales volume of the agricultural products in the current month based on the sliding window algorithm to obtain the real-time sales volume of the agricultural products in the current month;

[0130] Specifically, the calculation formula of the sliding window algorithm is as follows:

[0131]

[0132] Wherein, represents the total number of days in the current month, represents the total number of days for collecting the sales volume data, that is, represents the size of the window, represents the adjustment coefficient, represents rounding down, represents the minimum number of days for data collection, represents the maximum number of days for data collection.

[0133] Specifically, the calculation formula of the adjustment coefficient is as follows:

[0134]

[0135] Wherein, represents a random number, The calculation formula of is as follows;

[0136]

[0137] Wherein, represents the calculation formula of the linear congruential generator, represents the multiplication factor, represents the nth generated pseudo-random number, represents the increment, represents the modulus, and , .

[0138] For example, given an initial seed , by continuously iterating the above formula, a pseudo-random number sequence 、 ..... can be generated. The generated pseudo-random number ranges from 0 to , is the random floating-point number between 0 and 1 generated at the nth time.

[0139] Specifically, the calculation formula of the real-time sales volume is as follows:

[0140]

[0141] Wherein, represents the real-time sales volume on the first day of the current month, represents the real-time sales volume on the second day of the current month, represents the The real-time sales volume of the day, represents the real-time sales volume.

[0142] Specifically, The value range of is expressed as , by setting , it can ensure that the collected sales volume data has a sufficient time span, avoiding the prediction instability or misleading caused by too little collection of the sales volume data, ensuring that the sliding window can collect the sales volume data within a certain period, and capturing long-term trends such as seasonal changes and periodic fluctuations; by setting , it can avoid the influence of an overly long sliding window on the update of the importance score. For example, if the number of days for collecting the sales volume data is too long, the early information in the historical data may dilute the recent dynamic changes in the market, thereby reducing the response ability of the importance score update formula to the latest trends.

[0143] Update the importance score of the target word based on the predicted sales volume and the real-time sales volume, where the update formula for the importance score is as follows:

[0144]

[0145] Among them, represents the importance score of the target word, represents the real-time sales volume, represents the predicted sales volume, represents the total number of days in the current month, represents the total number of days for collecting the sales volume data, represents the promotion indicator variable, represents the correction coefficient of the real-time sales volume error, represents the correction coefficient of the influence of the promotion activity, represents the updated importance score;

[0146] When , then increase the importance score of the target word, otherwise decrease it, downgrade the target word, and re-rate it as a keyword. Perform a bubble sort on the importance scores of this keyword and the importance scores of other keywords. The value of the importance score of the keyword ranked first is the largest, and the value of the importance score of the keyword ranked last is the smallest. Select the keywords corresponding to the top 3 importance scores, and record these 3 keywords as value tags.

[0147] Value tags are keywords that are the most valuable and can best reflect current market trends or demands under current market conditions. By dynamically evaluating the importance of keywords, value tags ensure that agricultural product e-commerce platforms can respond to market changes in a timely manner, optimize agricultural product recommendations, pricing strategies, and marketing activities. This method can effectively avoid the lag effect, enabling agricultural product e-commerce platforms to more accurately meet consumer needs and improve the customer experience. At the same time, value tags also help agricultural product e-commerce platforms quickly adapt to market fluctuations, enhance competitiveness, and ultimately drive sales growth and profitability improvement.

[0148] Specifically, can randomly take values between This effectively avoids the situation where the updated importance score fluctuates violently due to abnormal data appearing in a certain fixed window. In this way, the update formula of the importance score becomes more robust, successfully avoiding the interference of data extreme values on the update formula of the importance score and ensuring the stability and reliability of the update process of the importance score.

[0149] As Figure 2 shown, it is a functional module diagram of a data value analysis system based on machine learning provided by an embodiment of the present invention.

[0150] The data value analysis system 100 based on machine learning described in the present invention can be installed in an electronic device. According to the functions implemented, the data value analysis system 100 based on machine learning can include a collection module 101, a keyword group generation module 102, a purchase intention combination generation module 103, an importance score generation module 104, a screening module 105, a predicted sales volume module 106, and a value tag generation module 107. The modules described in the present invention can also be referred to as units, which refer to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, and are stored in the memory of the electronic device.

[0151] In this embodiment, the functions of each module / unit are as follows:

[0152] The collection module is used to collect customer data of the agricultural product e-commerce platform;

[0153] The keyword group generation module is used to generate a keyword group of the customer according to the customer data;

[0154] The purchase intention combination generation module is used to generate the purchase intention combination of the customer based on the decision tree model and the keyword group;

[0155] The importance score generation module is used to perform importance analysis on the keyword group according to the purchase intention combination to obtain the importance score of the keywords in the keyword group;

[0156] The screening module is configured to screen out target words from the keywords according to the screening mechanism of the importance score;

[0157] The predicted sales volume module is configured to generate the predicted sales volume of agricultural products according to the target words and the customer data;

[0158] The value label generation module is configured to monitor the real-time sales volume of the agricultural products, update the importance score of the target words based on the predicted sales volume and the real-time sales volume, and update the target words according to the updated importance score to obtain the value label of the agricultural products.

[0159] In several embodiments provided by the present invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0160] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0161] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware, or in the form of a combination of hardware and software functional modules.

[0162] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0163] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence is a theory, method, technology and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A data value analysis method based on machine learning, characterized in that: The method comprises: Collect customer data of agricultural product e-commerce platforms; generating a keyword group for a customer based on the customer data; Generating the customer's purchase intention combination based on the decision tree model and the keyword group; Performing importance analysis on the keyword group according to the purchase intention combination to obtain importance scores of the keywords in the keyword group; Filtering out target words from the keywords according to the importance score screening mechanism; generating predicted sales of agricultural products based on the target words and the customer data; The real-time sales volume of the agricultural product is monitored, and the importance score of the target word is updated based on the predicted sales volume and the real-time sales volume, wherein the update formula of the importance score is as follows: ； in, Indicates the first Keywords, represents the importance score of the target word, Indicates real-time sales volume. It indicates the predicted sales volume. Indicates the total number of days in the month. Indicates the total number of days for collecting the sales data. represents the promotion indicator variable, Indicates the correction coefficient of real-time sales error, The correction factor representing the impact of promotional activities, represents the updated importance score; The target word is updated according to the updated importance score to obtain a value label for the agricultural product.

2. A data value analysis method based on machine learning as claimed in claim 1, characterized in that: The generating the customer's purchase intention combination based on the decision tree model and the keyword group includes: T1. Generate a data set of the customer according to the keywords and the customer data, wherein the keywords include: agricultural product A, agricultural product B and agricultural product C, and the customer data includes: purchase history, browsing time and browsing times of agricultural products; T2, inputting the data set into a decision tree model; T3. When the keyword has a purchase history, the keyword is marked as Y. When the keyword does not have a purchase history, the browsing time judgment algorithm is entered, wherein the browsing time judgment algorithm is: When the browsing time of the keyword When , the keyword is marked as Y; When the browsing time of the keyword is , then enter the browsing count judgment algorithm, wherein the browsing count judgment algorithm is: when the browsing count of the keyword times, the keyword is marked as Y. times, marking the keyword as N; When the browsing time of the keyword When , the keyword is marked as N; T4. Add the keyword marked as Y to the customer's purchase intention combination.

3. The data value analysis method based on machine learning according to claim 1, characterized in that: The step of performing importance analysis on the keyword group according to the purchase intention combination to obtain importance scores of the keywords in the keyword group includes: The purchase history impact score, browsing time impact score and browsing times impact score of the keywords in the keyword group are generated according to the purchase intention combination, wherein: The calculation formula of purchase history impact score is as follows: ； in, Indicates the first Keywords, Indicates that purchase history affects the score. Indicates the customer, Indicates the purchase history of the keyword. Indicates that the customer has a purchase history of the keyword. Indicates that the customer has no purchase history for the keyword; The calculation formula for browsing time impact score is as follows: ； in, Indicates the first Keywords, Indicates that browsing time affects the rating. Indicates the browsing time of the keyword; The calculation formula for the impact of the number of views on the score is as follows: ； in, Indicates the first Keywords, Indicates that the number of views affects the rating. Indicates the number of views of the keyword; The purchase history impact score, the browsing time impact score, and the browsing times impact score are weighted and summed to obtain the importance score of the keyword, wherein the calculation formula for the weighted sum is as follows: ； in, Indicates the first Keywords, represents the importance score of the keyword, , , They respectively represent the weight coefficients of the purchase history affecting the score, the browsing time affecting the score, and the browsing times affecting the score.

4. The data value analysis method based on machine learning according to claim 1, characterized in that: The screening mechanism is as follows: ； in, Indicates the first Keywords, Indicates the preset importance threshold, Represents the target word after filtering, represents the importance score.

5. The data value analysis method based on machine learning according to claim 1, characterized in that: The generating the predicted sales volume of agricultural products according to the target word and the customer data includes: Substitute the target word and the customer data into a sales forecasting algorithm to obtain the predicted sales of agricultural products, wherein the sales forecasting algorithm is as follows: ； in, Indicates the first Keywords, Represents the word vector corresponding to the target word, represents the data vector corresponding to the customer data, Indicates The predicted sales volume corresponding to the keywords, Represents a mapping function.

6. The data value analysis method based on machine learning according to claim 1, characterized in that: Monitor the real-time sales of the agricultural products, including: The sales data of the agricultural product for the current month is collected based on a sliding window algorithm to obtain the real-time sales data of the agricultural product for the current month.

7. The data value analysis method based on machine learning according to claim 6, characterized in that: The calculation formula of the sliding window algorithm is as follows: ； in, Indicates the total number of days in the month. Indicates the total number of days for collecting the sales data. represents the adjustment factor, represents rounding down, Indicates the minimum number of days for data collection. Indicates the maximum number of days for data collection.

8. The data value analysis method based on machine learning according to claim 7, characterized in that: The calculation formula of the adjustment coefficient is as follows: ； in, represents the adjustment factor, Represents a random number.

9. The data value analysis method based on machine learning according to claim 6, characterized in that: The calculation formula of the real-time sales volume is as follows: ； in, Indicates the real-time sales volume on the first day of the month. Indicates the real-time sales volume on the second day of the month. Indicates the month The real-time sales volume of the day, Indicates real-time sales.

10. A data value analysis system based on machine learning, characterized in that: The system comprises: The collection module is used to collect customer data of agricultural product e-commerce platforms; A keyword group generating module, used for generating a keyword group for a customer according to the customer data; A purchase intention combination generating module, used for generating the customer's purchase intention combination based on a decision tree model and the keyword group; An importance score generating module, used for performing importance analysis on the keyword group according to the purchase intention combination to obtain importance scores of the keywords in the keyword group; A screening module, used to screen out target words from the keywords according to the screening mechanism of the importance score; A sales forecasting module, used for generating a forecasted sales volume of agricultural products according to the target words and the customer data; The value tag generation module is used to monitor the real-time sales volume of the agricultural product, and update the importance score of the target word based on the predicted sales volume and the real-time sales volume, wherein the update formula of the importance score is as follows: ； in, Indicates the first Keywords, represents the importance score of the target word, Indicates real-time sales volume. It indicates the predicted sales volume. Indicates the total number of days in the month. Indicates the total number of days for collecting the sales data. represents the promotion indicator variable, Indicates the correction coefficient of real-time sales error, The correction factor representing the impact of promotional activities, represents the updated importance score; The target word is updated according to the updated importance score to obtain a value label for the agricultural product.

Images