Medical enterprise network marketing management platform and method based on big data

Abstract

The invention provides a medical enterprise network marketing management platform and method based on big data, and particularly relates to the technical field of marketing management. The medical enterprise network marketing management method based on big data comprises the steps of collecting historical marketing data, and constructing a first marketing data set corresponding to each medical product type; the sales volume difference feature of each first marketing data set is extracted, and the sales volume difference feature of each first marketing data set is generated; clustering the plurality of first marketing data sets to generate a plurality of second marketing data sets; constructing a training data set and training the sales volume prediction model; and obtaining target marketing data in the target time period and processing the target marketing data through the sales volume prediction model to generate a sales volume prediction result of the target medicine product type. According to the invention, accurate prediction of the sales volume of the medicine products is realized through the sales volume prediction model, so that a medicine enterprise is helped to better insight into the market and make a strategy, and development of the medicine enterprise is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of marketing management, and particularly to a network marketing management platform and method for pharmaceutical enterprises based on big data. Background Art

[0002] With the rapid development of technology, the pharmaceutical industry is also constantly transforming towards digitalization and intelligentization. In this process, the importance of the network marketing management platform has become increasingly prominent. However, there are a series of problems with traditional pharmaceutical sales forecasting methods, which limit enterprises from achieving better performance in the highly competitive market. One of the main problems is the lack of accuracy of traditional methods. In the past, pharmaceutical enterprises usually relied on historical sales data, market research, and expert judgment to predict sales. However, these methods often have difficulty accurately reflecting the dynamic changes in the market and the changes in patient needs. Especially in the case of a large amount of messy data, the accuracy of these traditional methods is severely challenged. Traditional models often have difficulty effectively processing complex data relationships, resulting in a low accuracy of the trained models, thus affecting the formulation and implementation of sales strategies.

[0003] Therefore, there is an urgent need for a network marketing management platform and method for pharmaceutical enterprises based on big data to make full use of big data technology and advanced data analysis algorithms to improve the accuracy of sales forecasting and help enterprises better understand the market. Summary of the Invention

[0004] To solve the above technical problems, the present invention proposes a network marketing management platform and method for pharmaceutical enterprises based on big data, combining big data technology and deep learning technology to improve the accuracy of pharmaceutical product sales forecasting.

[0005] The first aspect of the present invention provides a network marketing management method for pharmaceutical enterprises based on big data, including:

[0006] Collect historical marketing data, classify the historical marketing data based on pharmaceutical product types, and construct a first marketing data set corresponding to each pharmaceutical product type;

[0007] Extract the sales time series feature data of each first marketing data set, and extract the sales difference features of each first marketing data set based on the sales time series feature data;

[0008] Traverse the sales difference features of each first marketing data set with a preset sliding window to generate the sales difference features of each first marketing data set;

[0009] Cluster multiple first marketing data sets based on sales volume difference features to generate multiple second marketing data sets, extract the sales volume difference features of each second marketing data set, and determine the fluctuation period of each second marketing data set according to the sales volume difference features;

[0010] Extract the sales volume time series feature data, environmental time series feature data, and periodic feature data of the fluctuation period of each second marketing data set, construct a training data set, and train a sales volume prediction model through the training data set to obtain a trained sales volume prediction model;

[0011] Obtain the target marketing data within the target period, and extract the target sales volume time series feature data, target environmental time series feature data, and target periodic feature data of the target pharmaceutical product type from the target marketing data;

[0012] Input the target sales volume time series feature data, target environmental time series feature data, and target periodic feature data of the target pharmaceutical product type into the sales volume prediction model to generate the sales volume prediction result of the target pharmaceutical product type.

[0013] Preferably, extracting the sales volume difference features of each first marketing data set based on the sales volume time series feature data includes:

[0014] The sales volume time series feature data includes the sales volume data corresponding to a plurality of consecutive reference periods. Traverse the sales volume time series feature data, calculate the difference between the sales volume data corresponding to every two adjacent reference periods, and generate the sales volume difference features of each first marketing data set.

[0015] Preferably, traverse the sales volume difference features of each first marketing data set with a preset sliding window to generate the sales volume difference features of each first marketing data set, including:

[0016] For any sales volume difference feature, traverse the sales volume difference feature with a preset sliding window based on a preset step size. For any sliding operation, perform a summation operation on all elements within the preset sliding window to generate a difference factor corresponding to each preset sliding window, and generate the sales volume difference features corresponding to the sales volume difference features of each first marketing data set according to multiple difference factors.

[0017] Preferably, clustering multiple first marketing data sets based on the sales volume difference features to generate multiple second marketing data sets includes:

[0018] Normalize each sales volume difference feature, including recording the difference factors greater than 0 in the sales volume difference feature as the first eigenvalue, recording the difference factors less than 0 in the sales volume difference feature as the second eigenvalue, generating a sales volume difference vector for each sales volume difference feature, calculating the similarity between any two sales volume difference vectors, and clustering multiple first marketing data sets according to multiple similarities to generate multiple second marketing data sets;

[0019] Among them, clustering multiple first marketing data sets according to multiple similarities to generate multiple second marketing data sets includes:

[0020] S11. For the i-th first marketing data set Li, let i = 1;

[0021] S12. Construct a second marketing data set based on the first marketing data set Li;

[0022] S13. Screen out the first marketing data sets whose similarity to the sales volume difference vector of the first marketing data set Li is greater than the preset similarity threshold, sort all the screened first marketing data sets according to the size of the similarity, traverse each screened first marketing data set, and judge whether the similarity between the sales volume difference vector of the first marketing data set and the sales volume difference vectors of all the first marketing data sets in the second marketing data set to which the first marketing data set Li belongs is greater than the preset similarity threshold. If so, write the first marketing data set into the second marketing data set to which the first marketing data set Li belongs and no longer regard it as a first marketing data set;

[0023] S14. Judge whether i < n holds, where n is the number of initial first marketing data sets. If it holds, let i = i + 1 and go to step S15; otherwise, go to step S16;

[0024] S15. Judge whether the i-th first marketing data set exists. If it exists, go to step S12; otherwise, go to step S14;

[0025] S16. Output all the second marketing data sets.

[0026] Preferably, determining the fluctuation period of each second marketing data set according to the sales volume difference feature includes:

[0027] Traverse the sales volume difference features of each second marketing data set. For the sales volume difference feature of any second marketing data set, for the element value of each reference period in the sales volume difference feature, if the absolute value of the element value of the reference period is greater than the preset difference threshold, record the reference period as the fluctuation period to generate the fluctuation period of each second marketing data set.

[0028] Preferably, for the sales volume prediction model, it further includes:

[0029] Taking the environmental time series feature data and periodic feature data in the training dataset as inputs, and taking the sales volume time series feature data in the training dataset as the training target, training the sales volume prediction model, where the sales volume prediction model is an LSTM network model;

[0030] The sales volume prediction model includes multiple prediction units, and the sales volume time series feature data, environmental time series feature data, and periodic feature data extracted from each second marketing data set are respectively used for the training of one of the prediction units.

[0031] The second aspect of the present invention provides a network marketing management platform for pharmaceutical enterprises based on big data, which is used to implement the above-mentioned network marketing management method for pharmaceutical enterprises based on big data, and includes:

[0032] A data collection module, which is used to collect historical marketing data;

[0033] A data classification module, which is used to classify historical marketing data based on the types of pharmaceutical products, and construct a first marketing data set corresponding to each type of pharmaceutical product;

[0034] A differential feature extraction module, which is used to extract the sales volume time series feature data of each first marketing data set, and extract the sales volume differential feature of each first marketing data set based on the sales volume time series feature data;

[0035] A data clustering module, which is used to traverse the sales volume differential features of each first marketing data set with a preset sliding window, generate the sales volume difference features of each first marketing data set, and cluster multiple first marketing data sets based on the sales volume difference features to generate multiple second marketing data sets;

[0036] A sample construction module, which is used to extract the sales volume differential features of each second marketing data set, determine the fluctuation period of each second marketing data set according to the sales volume differential features, and extract the sales volume time series feature data, environmental time series feature data, and periodic feature data of the fluctuation period of each second marketing data set to construct a training data set;

[0037] A model training module, which is used to train the sales volume prediction model through the training data set to obtain a trained sales volume prediction model;

[0038] A sales prediction module, which is used to obtain target marketing data within a target time period, extract target sales time series feature data, target environmental time series feature data, and target cycle feature data of a target pharmaceutical product type from the target marketing data, and input the target sales time series feature data, target environmental time series feature data, and target cycle feature data of the target pharmaceutical product type into a sales prediction model to generate a sales prediction result of the target pharmaceutical product type.

[0039] Preferably, for the differential feature extraction module, sales differential features of each first marketing data set are extracted based on the sales time series feature data, including:

[0040] Traverse the sales time series feature data, calculate the difference between the sales data corresponding to every two adjacent reference time periods, and generate the sales differential features of each first marketing data set.

[0041] The present invention has the following beneficial effects:

[0042] 1. The present invention collects historical marketing data of pharmaceutical enterprises through big data technology. For the collected historical marketing data, a first marketing data set corresponding to each pharmaceutical product type is constructed, differential feature analysis is performed on each first marketing data set, sales difference features of each first marketing data set are extracted, and multiple first marketing data sets are clustered to obtain multiple second marketing data sets, realizing dimensionality reduction of the data. A training data set is constructed based on multiple second marketing data sets and the sales prediction model is trained. The sales of pharmaceutical products are accurately predicted through the sales prediction model to help pharmaceutical enterprises better understand the market and formulate strategies, promoting the development of pharmaceutical enterprises. Description of the Drawings

[0043] Figure 1 It is a schematic flowchart of a network marketing management method for pharmaceutical enterprises based on big data provided by an embodiment of the present invention. Detailed Embodiments

[0044] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. 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.

[0045] See Figure 1 , a network marketing management method for pharmaceutical enterprises based on big data provided by an embodiment of the present invention specifically includes the following steps:

[0046] S1. Collect historical marketing data, classify the historical marketing data based on the types of pharmaceutical products, construct a first marketing data set corresponding to each type of pharmaceutical product, extract the sales time-series feature data of each first marketing data set, and extract the sales difference features of each first marketing data set based on the sales time-series feature data.

[0047] Specifically, collect the historical marketing data of pharmaceutical enterprises through big data technology. Among them, the historical marketing data records the relevant sales data of different pharmaceutical products sold by pharmaceutical enterprises, such as the sales volume data of different products in different periods and the environmental weather data in different periods. For the convenience of subsequent data processing, divide the historical marketing data based on the type differences of pharmaceutical products. Exemplarily, divide pharmaceutical products according to the treatment object, which can be divided into multiple major categories of pharmaceutical products for treating the cardiovascular system, respiratory system, digestive system, nervous system, etc. Each major category includes multiple sub-categories. In this embodiment, classify the historical marketing data based on the specific types of drugs sold by pharmaceutical enterprises, and construct a first marketing data set corresponding to each type of pharmaceutical product, that is, each drug corresponds to a first marketing data set.

[0048] In this embodiment, for each first marketing data set, extract the sales time-series feature data of each first marketing data set. Among them, the sales time-series feature data includes the sales volume data corresponding to a continuous number of reference periods. Exemplarily, taking days as the basic unit and the marketing data in the past year as the analysis object, the sales time-series feature data represents the daily sales volume data of a certain pharmaceutical product in the past year. Those skilled in the art can also set the reference period according to the actual situation, such as 3 days, 7 days, etc. This embodiment does not specifically limit it.

[0049] After extracting the sales time-series feature data of each first marketing data set, perform differential processing on each group of sales time-series feature data to generate the sales difference features of each first marketing data set.

[0050] Specifically, taking the sales time-series feature data of any one first marketing data set as an example, traverse this group of sales time-series feature data, calculate the difference between the sales volume data corresponding to every two adjacent reference periods, and generate the sales difference features of each first marketing data set. The sales difference features are used to characterize the difference in sales volume between each reference period and the previous reference period among multiple reference periods.

[0051] S2. Traverse the sales volume difference features of each first marketing data set with a preset sliding window to generate the sales volume difference features of each first marketing data set. Cluster multiple first marketing data sets based on the sales volume difference features to generate multiple second marketing data sets. Extract the sales volume difference features of each second marketing data set, and determine the fluctuation period of each second marketing data set according to the sales volume difference features.

[0052] Specifically, the preset sliding window includes at least two reference periods. Exemplarily, assuming the reference period is 1 day, the preset sliding window can be 2 days, 3 days, 7 days, etc. In this embodiment, taking 3 days as an example, traverse the sales volume difference features with the preset sliding window based on the preset step size, and each sliding operation generates a difference factor.

[0053] In this embodiment, taking any sales volume difference feature as an example, traverse the sales volume difference features with the preset sliding window based on the preset step size. Taking one reference period as an example for the preset step size, for each sliding operation, perform a summation operation on all elements within the preset sliding window to generate the difference factor corresponding to each preset sliding window, and obtain the sales volume difference features including multiple difference factors. Perform the above operations on each first marketing data set to generate the sales volume difference features corresponding to the sales volume difference features of each first marketing data set. The sales volume difference features are used to represent the law of the sales volume of pharmaceutical products changing over time. The data volume is reduced through the sliding operation of the preset sliding window, and the overall change trend of the sales volume within the multiple reference periods corresponding to the preset sliding window is characterized by the difference factor.

[0054] In this embodiment, multiple first marketing data sets are also clustered through the sales volume difference features. Specifically, for each pharmaceutical product, the increasing and decreasing trends of its sales volume in different periods are considered, and the pharmaceutical products with relatively close change trends are clustered into a whole to improve the subsequent model training speed and accuracy. Multiple first marketing data sets are clustered to generate multiple second marketing data sets. By extracting and analyzing the sales volume difference features of each second marketing data set, the fluctuation period of each second marketing data set is determined, where the fluctuation period indicates that the sales volume of the pharmaceutical product has a certain increase or decrease during this period.

[0055] S3. Extract the sales volume time series feature data, environmental time series feature data, and periodic feature data of the fluctuation period of each second marketing data set, construct a training data set, and train the sales volume prediction model through the training data set to obtain a trained sales volume prediction model.

[0056] In this embodiment, for the training data set, it includes multiple sample subsets, and each sample subset corresponds to a second marketing data set, including sales time series feature data, environmental time series feature data, and periodic feature data of the fluctuation period of the second marketing data set. For the sales prediction model, the sales prediction model includes multiple prediction units.

[0057] During the process of model training, each sample in the training data set is used for training one of the prediction units of the sales prediction model. Specifically, for any one prediction unit, taking the sample subset corresponding to this prediction unit as an example, the environmental time series feature data and periodic feature data of this sample subset in the training data set are used as inputs, and the sales time series feature data of this sample subset is used as the training target to train this prediction unit in the sales prediction model. After completing the training of each prediction unit, a trained sales prediction model is obtained.

[0058] In this embodiment, the environmental time series feature data includes weather data on different dates, such as parameters like temperature, humidity, and light intensity. Affected by the environment, there may be some infectious diseases with seasonal characteristics in different periods. Therefore, the environmental time series feature data is selected as one of the data analysis objects for model training. The periodic feature data represents the periodicity of time. Exemplarily, features such as the week, month, and season to which the reference period belongs. For example, divided by season - month - week, the third day of the second week of March can be represented as 1 - 3 - 2 - 3, that is, the third day of the second week of the third month of the first season. Combining the periodic feature and the environmental feature for comprehensive analysis to deeply analyze the change rules of the sales volume of different pharmaceutical products.

[0059] In this embodiment, the sales prediction model is specifically a neural network model with time series data processing capabilities. In this embodiment, taking the LSTM network model as an example, a sales prediction model is constructed based on the LSTM network framework, which is used to train a model that can predict the sales volume of pharmaceutical products by analyzing environmental feature data and periodic feature data.

[0060] S4. Obtain the target marketing data within the target period, extract the target sales time series feature data, target environmental time series feature data, and target periodic feature data of the target pharmaceutical product type from the target marketing data and input them into the sales prediction model to generate the sales prediction result of the target pharmaceutical product type.

[0061] In this embodiment, the target time period can be reasonably set according to the actual situation. Exemplarily, if it is necessary to predict the sales volume of a certain pharmaceutical product in the next three days, a historical week is used as the target time period, and the target marketing data of this pharmaceutical product within 1 week before the current time is selected, including the daily sales volume data, environmental data, periodic characteristic data of the time period, etc. within 1 week before the current time, and input into the sales volume prediction model to generate the sales volume prediction result of this pharmaceutical product.

[0062] Furthermore, in step S2, to determine the fluctuation time period of each second marketing data set according to the sales volume difference characteristics, it specifically includes:

[0063] Traverse the sales volume difference characteristics of each second marketing data set. Taking the sales volume difference characteristics of any one second marketing data set as an example, for the element value of each reference time period in the sales volume difference characteristics, if the absolute value of the element value of the reference time period is greater than the preset difference threshold, then the reference time period is recorded as the fluctuation time period, and the fluctuation time period of each second marketing data set is generated.

[0064] In this embodiment, the element value of each reference time period represents the sales volume difference between the previous and the next day. The larger the absolute value of the element value of the reference time period, the greater the sales volume difference of a certain pharmaceutical product between the previous and the next day. By setting a reasonable preset difference threshold, the reference time period with large sales volume data fluctuations is recorded as the fluctuation time period, which is used to indicate the period with a large upward or downward trend of sales volume data among multiple reference time periods.

[0065] Furthermore, in step S2, to cluster multiple first marketing data sets based on the sales volume difference characteristics to generate multiple second marketing data sets, it specifically includes:

[0066] Perform normalization processing on each sales volume difference characteristic, including recording the difference factor greater than 0 in the sales volume difference characteristic as the first characteristic value, and recording the difference factor less than 0 in the sales volume difference characteristic as the second characteristic value, generating the sales volume difference vector of each sales volume difference characteristic, calculating the similarity of any two sales volume difference vectors, and clustering multiple first marketing data sets according to multiple similarities to generate multiple second marketing data sets.

[0067] In this embodiment, by normalizing each sales volume difference feature, similarity analysis can be better performed. The purpose of clustering multiple first marketing data sets is to determine two or more products among multiple pharmaceutical products whose periods of sales volume increase and decrease are relatively similar, and regard these products as a type, which is convenient for reducing the training difficulty of the model. Although training a prediction unit for each type of pharmaceutical product can better improve the prediction accuracy, many pharmaceutical products have relatively close sales volumes. For example, there are multiple pharmaceutical products for treating colds. During the peak cold season, the sales volumes of these multiple pharmaceutical products will all show an upward trend. Therefore, with reference to the increase or decrease in sales volume, the difference factors greater than 0 in the sales volume difference feature are recorded as the first feature values, and the difference factors less than 0 in the sales volume difference feature are recorded as the second feature values. Exemplarily, they are respectively recorded as 1 and -1, so as to realize the normalization of the sales volume difference feature. Based on the normalized sales volume difference feature, a corresponding sales volume difference vector is generated, that is, a one-dimensional vector composed of three values of -1, 0, and 1. The similarity between any two sales volume difference vectors is measured by calculating the similarity between each two sales volume difference vectors, and then multiple first marketing data sets are clustered according to multiple similarities to generate multiple second marketing data sets. Among them, the calculation method of similarity is a technique well-known to those skilled in the art, such as Euclidean distance, cosine similarity, Jaccard similarity coefficient, etc. In this embodiment, cosine similarity is taken as an example to calculate the similarity.

[0068] In this embodiment, after calculating the similarity between any two sales volume difference vectors, multiple first marketing data sets are clustered according to multiple similarities to generate multiple second marketing data sets, which specifically includes:

[0069] S11. For the i-th first marketing data set Li, let i = 1;

[0070] S12. Construct a second marketing data set based on the first marketing data set Li;

[0071] S13. Screen out the first marketing data sets whose similarity with the sales volume difference vector of the first marketing data set Li is greater than the preset similarity threshold, sort all the screened first marketing data sets according to the size of the similarity, traverse each screened first marketing data set, and judge whether the sales volume difference vector of the first marketing data set is greater than the preset similarity threshold with the sales volume difference vectors of all the first marketing data sets in the second marketing data set to which the first marketing data set Li belongs. If so, write the first marketing data set into the second marketing data set to which the first marketing data set Li belongs and no longer regard it as a first marketing data set;

[0072] It should be noted that it is easy to measure the similarity between any two first marketing data sets by presetting a similarity threshold. For further clustering and reducing the data volume, after pairwise clustering is achieved, clustering among multiple first marketing data sets is realized through this step. In each finally generated second marketing data set, the similarity between any two first marketing data sets is greater than the preset similarity threshold, ensuring the accuracy of clustering and improving the quality of the training data set used for model training.

[0073] S14. Determine whether i is less than n, where n is the number of initial first marketing data sets. If it holds, let i = i + 1 and go to step S15; otherwise, go to step S16.

[0074] S15. Determine whether the i-th first marketing data set exists. If it exists, go to step S12; otherwise, go to step S14.

[0075] S16. Output all second marketing data sets.

[0076] Through the above steps, all first marketing data sets are processed, and multiple second marketing data sets are obtained by clustering.

[0077] A method for network marketing management of pharmaceutical enterprises based on big data provided by an embodiment of the present invention collects historical marketing data of pharmaceutical enterprises through big data technology. For the collected historical marketing data, a first marketing data set corresponding to each pharmaceutical product type is constructed. Differential feature analysis is performed on each first marketing data set, the sales volume difference features of each first marketing data set are extracted, and multiple first marketing data sets are clustered to obtain multiple second marketing data sets, realizing dimensionality reduction of the data. A training data set is constructed based on multiple second marketing data sets, and a sales volume prediction model is trained. The sales volume of pharmaceutical products is accurately predicted through the sales volume prediction model to help pharmaceutical enterprises better understand the market and formulate strategies, promoting the development of pharmaceutical enterprises.

[0078] The present invention also provides a network marketing management platform for pharmaceutical enterprises based on big data, which is used to implement the above-mentioned method for network marketing management of pharmaceutical enterprises based on big data, and specifically includes:

[0079] A data collection module, which is used to collect historical marketing data based on big data technology;

[0080] A data classification module, which is used to classify historical marketing data based on pharmaceutical product types and construct a first marketing data set corresponding to each pharmaceutical product type;

[0081] The differential feature extraction module is used to extract the sales time series feature data of each first marketing data set, and extract the sales differential features of each first marketing data set based on the sales time series feature data;

[0082] Specifically, traverse the sales time series feature data, calculate the difference between the sales data corresponding to every two adjacent reference periods, and generate the sales differential features of each first marketing data set.

[0083] The data clustering module is used to traverse the sales differential features of each first marketing data set with a preset sliding window, generate the sales difference features of each first marketing data set, and cluster multiple first marketing data sets based on the sales difference features to generate multiple second marketing data sets;

[0084] Specifically, for any sales differential feature, traverse the sales differential feature with a preset sliding window based on a preset step size. For any sliding operation, perform a summation operation on all elements within the preset sliding window to generate a difference factor corresponding to each preset sliding window, and generate the sales difference features corresponding to the sales differential features of each first marketing data set according to multiple difference factors.

[0085] The sample construction module is used to extract the sales differential features of each second marketing data set, determine the fluctuation periods of each second marketing data set according to the sales differential features, extract the sales time series feature data, environmental time series feature data, and periodic feature data of the fluctuation periods of each second marketing data set, and construct a training data set;

[0086] The model training module is used to train the sales prediction model with the training data set to obtain a trained sales prediction model;

[0087] Specifically, use the environmental time series feature data and periodic feature data in the training data set as inputs, and use the sales time series feature data in the training data set as the training target to perform model training on the sales prediction model. The sales prediction model is an LSTM network model;

[0088] The sales prediction model includes multiple prediction units, and the sales time series feature data, environmental time series feature data, and periodic feature data extracted from each second marketing data set are respectively used for the training of one of the prediction units.

[0089] The sales prediction module is used to obtain the target marketing data within the target period, extract the target sales time series feature data, target environmental time series feature data, and target periodic feature data of the target pharmaceutical product type from the target marketing data, and input the target sales time series feature data, target environmental time series feature data, and target periodic feature data of the target pharmaceutical product type into the sales prediction model to generate the sales prediction result of the target pharmaceutical product type.

[0090] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The parts not described in detail in this specification belong to the prior art well known to those skilled in the art.

Claims

1. A network marketing management method for pharmaceutical enterprises based on big data, characterized in that, Including: Collect historical marketing data, classify the historical marketing data based on the types of pharmaceutical products, and construct a first marketing data set corresponding to each type of pharmaceutical product; Extract the sales time series feature data of each first marketing data set, and extract the sales difference features of each first marketing data set based on the sales time series feature data; Traverse the sales difference features of each first marketing data set with a preset sliding window to generate the sales difference features of each first marketing data set; Cluster multiple first marketing data sets based on the sales difference features to generate multiple second marketing data sets, extract the sales difference features of each second marketing data set, and determine the fluctuation period of each second marketing data set according to the sales difference features; Extract the sales time series feature data, environmental time series feature data, and periodic feature data of the fluctuation period of each second marketing data set, construct a training data set, and train a sales prediction model through the training data set to obtain a trained sales prediction model; Obtain the target marketing data within the target period, and extract the target sales time series feature data, target environmental time series feature data, and target periodic feature data of the target pharmaceutical product type from the target marketing data; Input the target sales time series feature data, target environmental time series feature data, and target periodic feature data of the target pharmaceutical product type into the sales prediction model to generate the sales prediction result of the target pharmaceutical product type.

2. The method for network marketing management of pharmaceutical enterprises based on big data according to claim 1, wherein, Extracting the sales difference features of each first marketing data set based on the sales time series feature data includes: The sales time series feature data includes the sales data corresponding to a plurality of consecutive reference periods. Traverse the sales time series feature data, calculate the difference between the sales data corresponding to every two adjacent reference periods, and generate the sales difference features of each first marketing data set.

3. The network marketing management method for pharmaceutical enterprises based on big data according to claim 1, characterized in that Traversing the sales difference features of each first marketing data set with a preset sliding window to generate the sales difference features of each first marketing data set includes: For any sales difference feature, traverse the sales difference feature with a preset sliding window based on a preset step length. For any sliding operation, perform a summation operation on all elements within the preset sliding window to generate a difference factor corresponding to each preset sliding window, and generate the sales difference features corresponding to the sales difference features of each first marketing data set according to multiple difference factors.

4. The method for network marketing management of pharmaceutical enterprises based on big data according to claim 3, wherein Clustering multiple first marketing data sets based on the sales difference features to generate multiple second marketing data sets includes: Perform normalization processing on each sales difference feature, including recording the difference factors greater than 0 in the sales difference feature as the first feature value, recording the difference factors less than 0 in the sales difference feature as the second feature value, generating a sales difference vector for each sales difference feature, calculating the similarity between any two sales difference vectors, and clustering multiple first marketing data sets according to multiple similarities to generate multiple second marketing data sets; Among them, clustering multiple first marketing data sets according to multiple similarities to generate multiple second marketing data sets includes: S11. For the i-th first marketing data set Li, let i = 1; S12. Construct a second marketing data set based on the first marketing data set Li; S13. Screen out the first marketing data sets whose similarity to the sales volume difference vector of the first marketing data set Li is greater than the preset similarity threshold. Sort all the screened first marketing data sets according to the size of the similarity. Traverse each screened first marketing data set and determine whether the similarity between the sales volume difference vector of the first marketing data set and the sales volume difference vectors of all the first marketing data sets in the second marketing data set to which the first marketing data set Li belongs is greater than the preset similarity threshold. If so, write the first marketing data set into the second marketing data set to which the first marketing data set Li belongs and no longer regard it as a first marketing data set; S14. Determine whether i < n holds, where n is the number of initial first marketing data sets. If it holds, let i = i + 1 and go to step S15; otherwise, go to step S16; S15. Determine whether the i-th first marketing data set exists. If it exists, go to step S12; otherwise, go to step S14; S16. Output all the second marketing data sets.

5. The method for network marketing management of pharmaceutical enterprises based on big data according to claim 2, characterized in that, Determine the fluctuation period of each second marketing data set according to the sales volume difference characteristics, including: Traverse the sales volume difference characteristics of each second marketing data set. For the sales volume difference characteristics of any one second marketing data set, for the element value of each reference period in the sales volume difference characteristics, if the absolute value of the element value of the reference period is greater than the preset difference threshold, then record the reference period as the fluctuation period to generate the fluctuation period of each second marketing data set.

6. The network marketing management method for pharmaceutical enterprises based on big data according to claim 1, characterized in that For the sales volume prediction model, it further includes: Use the environmental time series feature data and periodic feature data in the training data set as inputs, and use the sales volume time series feature data in the training data set as the training target to train the sales volume prediction model. The sales volume prediction model is an LSTM network model; The sales volume prediction model includes multiple prediction units. The sales volume time series feature data, environmental time series feature data, and periodic feature data extracted from each second marketing data set are respectively used for the training of one of the prediction units.

7. A network marketing management platform for pharmaceutical enterprises based on big data, which is used to implement the network marketing management method for pharmaceutical enterprises based on big data described in any one of the above claims 1-6, and is characterized in that, It includes: A data collection module for collecting historical marketing data; A data classification module for classifying historical marketing data based on the types of pharmaceutical products and constructing a first marketing data set corresponding to each type of pharmaceutical product; A difference feature extraction module for extracting the sales volume time series feature data of each first marketing data set and extracting the sales volume difference characteristics of each first marketing data set based on the sales volume time series feature data; A data clustering module for traversing the sales volume difference characteristics of each first marketing data set with a preset sliding window to generate the sales volume difference characteristics of each first marketing data set, and clustering multiple first marketing data sets based on the sales volume difference characteristics to generate multiple second marketing data sets; A sample construction module, which is used to extract the sales volume difference features of each second marketing data set, determine the fluctuation period of each second marketing data set according to the sales volume difference features, extract the sales volume time series feature data, environmental time series feature data and periodic feature data of the fluctuation period of each second marketing data set, and construct a training data set; A model training module, which is used to train the sales volume prediction model with the training data set to obtain a trained sales volume prediction model; A sales volume prediction module, which is used to obtain the target marketing data within the target period, extract the target sales volume time series feature data, target environmental time series feature data and target periodic feature data of the target pharmaceutical product type from the target marketing data, and input the target sales volume time series feature data, target environmental time series feature data and target periodic feature data of the target pharmaceutical product type into the sales volume prediction model to generate the sales volume prediction result of the target pharmaceutical product type.

8. The network marketing management platform for pharmaceutical enterprises based on big data according to claim 7, characterized in that For the difference feature extraction module, based on the sales volume time series feature data, extract the sales volume difference features of each first marketing data set, including: Traverse the sales volume time series feature data, calculate the difference between the sales volume data corresponding to every two adjacent reference periods, and generate the sales volume difference features of each first marketing data set.

Images