Report customization method and device based on medical data

Abstract

The embodiment of the invention provides a report customization method and device based on medical data. A report customization method based on medical data belongs to the technical field of data processing, and obtains standardized data after redundant data, missing data and inaccurate data are removed through data cleaning and standardization processing, so as to improve the data source quality of a generated report. The method comprises the following steps: standardizing data of a user demand report, generating structured data by performing multi-dimensional analysis and feature extraction on the standardized data to highlight the correlation between the data so as to facilitate the generation of a high-quality report, and finally, mapping the structured data to a user demand report template and combining compliance check and risk assessment rules embedded in the user demand report template so as to improve the user demand report quality. And the function that the report meets the customized demand of a customer is further improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a report customization method and device based on medical data. Background Art

[0002] As regulatory requirements for the pharmaceutical industry become increasingly stringent, data collection, management, and analysis are becoming increasingly crucial for corporate compliance, market decision-making, and resource allocation. In particular, the pharmaceutical industry lacks professional data management systems, making data operations complex and error-prone, making it difficult to effectively manage large and diverse data volumes.

[0003] To address this, companies generally consider presenting data through data reports. However, existing technologies have numerous issues with data processing efficiency, data quality consistency, report customization flexibility, and regulatory compliance, leading to significant difficulties for companies in report generation and decision support.

[0004] For example, existing report generation methods generally rely on manual processing, which compromises data accuracy and is prone to issues such as inaccurate data, inconsistent formats, and missing fields. Furthermore, traditional manual data processing is inefficient, resulting in high costs and time-consuming data processing. Furthermore, with increasingly stringent regulatory policies, existing report generation methods struggle to fully meet compliance requirements. Furthermore, report customization lacks flexibility, making it difficult to address diverse business needs and personalized requirements. Summary of the Invention

[0005] In view of this, an object of an embodiment of the present invention is to provide a report customization method and apparatus based on medical data to solve at least one of the above technical problems.

[0006] To achieve the above objectives, in a first aspect, the present invention provides a report customization method based on medical data, comprising:

[0007] Obtain pharmaceutical sales data from multiple pharmaceutical industry data sources;

[0008] Cleaning and standardizing the pharmaceutical sales data to obtain standardized data after removing redundant data, missing data, and inaccurate data;

[0009] Generating structured data by performing multi-dimensional analysis and feature extraction on the standardized data;

[0010] Based on matching the structured data with the template fields in the user demand report template, the structured data is mapped to the user demand report template, and combined with the compliance check and risk assessment rules embedded in the user demand report template, a personalized report is generated.

[0011] In a second aspect, a report customization device based on medical data is provided, comprising:

[0012] A data acquisition unit, used to acquire pharmaceutical sales data from multiple pharmaceutical industry data sources;

[0013] A data processing unit, configured to perform data cleaning and standardization on the pharmaceutical sales data to obtain standardized data after removing redundant data, missing data, and inaccurate data;

[0014] A feature extraction unit, configured to generate structured data by performing multi-dimensional analysis and feature extraction on the standardized data;

[0015] A report generation unit is used to match the structured data with the template fields in the user demand report template, map the structured data to the user demand report template, and generate a personalized report in combination with the compliance check and risk assessment rules embedded in the user demand report template.

[0016] The above technical solution has the following beneficial effects:

[0017] The present invention provides a report customization method based on pharmaceutical data, comprising: obtaining pharmaceutical sales data from multiple pharmaceutical industry data sources; performing data cleaning and standardization on the pharmaceutical sales data to obtain standardized data after removing redundant data, missing data, and inaccurate data; performing multi-dimensional analysis and feature extraction on the standardized data to generate structured data; matching the structured data with template fields in a user demand report template, mapping the structured data to the user demand report template, and generating a personalized report in combination with the compliance check and risk assessment rules embedded in the user demand report template. By performing data cleaning and standardization on the standardized data to obtain standardized data after removing redundant data, missing data, and inaccurate data, the quality of the data source for generating the report is improved; then, performing multi-dimensional analysis and feature extraction on the standardized data to generate structured data, the correlation between the data is highlighted, which is conducive to the generation of high-quality reports; finally, by mapping the structured data to the user demand report template, and combining the compliance check and risk assessment rules embedded in the user demand report template, the function of the report to meet customer customization needs is further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 is a flow chart of a method for customizing a report based on medical data according to an embodiment of the present invention;

[0020] Figure 2 is a schematic diagram of the data analysis background of an embodiment of the present invention;

[0021] Figure 3 This is a report output and optimization flow chart of an embodiment of the present invention;

[0022] Figure 4 This is a structural diagram of report customization according to an embodiment of the present invention;

[0023] Figure 5 This is a schematic diagram of the structure of a report generating device for medical data according to an embodiment of the present invention;

[0024] Figure 6 This is a functional block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0026] Example 1

[0027] The present disclosure provides a report customization method based on medical data. Figure 1 FIG. 1 is a flow chart of a report customization method based on medical data according to an exemplary embodiment. Figure 1 As shown, the report customization method based on medical data includes:

[0028] Step 10: Obtain pharmaceutical sales data from multiple pharmaceutical industry data sources;

[0029] Step 11: Clean and standardize the pharmaceutical sales data to obtain standardized data after removing redundant data, missing data, and inaccurate data;

[0030] Step 12: Generate structured data by performing multi-dimensional analysis and feature extraction on the standardized data;

[0031] Step 13: Based on the matching of the structured data with the template fields in the user demand report template, the structured data is mapped to the user demand report template, and a personalized report is generated in combination with the compliance check and risk assessment rules embedded in the user demand report template.

[0032] In this exemplary embodiment, Figure 2 Schematic diagram of the data analysis background of the embodiment of the present invention. Figure 2As shown, the report generation process generally includes four parts: omnichannel data collection, intelligent data cleaning, data appeals, and collaborative data delivery. Collection channels can cover graded hospitals, primary healthcare institutions, public health institutions, independent pharmacies, chain pharmacies, commercial companies, and e-commerce companies. Intelligent data cleaning includes terminal cleaning, product cleaning, unit conversion, and batch number cleaning. Terminal cleaning requires standardizing the format of raw data collected from different terminals (e.g., public health institutions and direct-to-consumer pharmacies) such as graded hospitals and primary healthcare institutions, correcting missing and duplicate fields caused by equipment differences or manual data entry. Product cleaning eliminates ambiguity by aligning product names across e-commerce platforms, chain pharmacies, and other channels through semantic analysis (e.g., mapping "Aspirin Enteric-coated Tablets" and "Aspirin Enteric" to a unified code). Unit conversion automatically converts dispersed data (e.g., "boxes" for commercial companies, "bottles" for independent pharmacies) into standard units based on preset rules. Batch number cleaning performs logical analysis and format normalization on batch numbers (e.g., "A21-3B" and "a21#3b") that are mixed across e-commerce and offline channels, supporting flow query and supply chain traceability. Data appeals include four components: organizational structure, flow query, appeal submission, and appeal review. Combined with the "data appeal" module shown in the figure, the process achieves closed-loop governance of abnormal data through four links: the organizational structure clarifies the rights and responsibilities of the supply chain, quality control and other departments within the enterprise, and provides responsible entities for data traceability; the flow query is based on the cleaned standardized data (such as product batch number, terminal information), supports cross-channel tracking of drug distribution paths (such as the specific flow from commercial companies to graded hospitals), and quickly locates abnormal nodes (such as inventory differences or sales breakpoints); the appeal allows relevant parties to submit problem data through the system (such as unit errors and batch number conflicts found during terminal cleaning), and associate the original documents with the cleaning records; the appeal review is conducted by the quality control department to review the content of the appeal and determine whether to trigger data rollback (such as returning to the "unit conversion" link for re-cleaning) or business correction. Based on the above data and channel experience, this application can adopt an automated data collection mechanism when acquiring multi-channel data, which can collect data in real time from different pharmaceutical industry data sources. At the same time, the data is pre-processed through the built-in data template fields and standardized modules to ensure data accuracy and consistency. Collaborative data delivery includes regular report analysis, customized report analysis and data source fusion. Collaborative data delivery refers to the process of collaboratively delivering data value by integrating data resources and needs from multiple parties. Conventional report analysis provides foundational support for daily decision-making through standardized periodic reports on sales and inventory, while customized report analysis conducts in-depth data mining and visualization based on the personalized needs of business scenarios (e.g., specific products, regions, or channels). Data source fusion, on the other hand, breaks down data silos across different collection channels (e.g., hospitals, pharmacies, e-commerce platforms), cleansing, aligning, and analyzing correlations across heterogeneous data, ultimately forming a holistic business analysis.The collaboration of the three has realized a closed loop of data services from standardization to customization, and from decentralization to unification, helping users to efficiently obtain multi-dimensional and feasible decision-making basis.

[0033] In this exemplary embodiment, while data collection is being performed, a data processing and analysis platform can be constructed to automatically optimize the data storage and reading process, thereby improving data processing efficiency.

[0034] In this exemplary embodiment, available data sources can be identified through an intelligent data source identification module. This module can identify data sources from different pharmaceutical industries, including but not limited to commercial companies, chain drug stores, DTP (Direct to Patient Pharmacy) drug stores, etc. The intelligent data source identification module selects different data sources for real-time collection based on the needs of the enterprise and specific scenarios. Each data source is marked as a valid data source, and the priority of data collection is determined based on its characteristics (such as data format, update frequency, etc.). In this way, through the automated data source identification and selection process, it is possible to flexibly select suitable data sources according to the needs of the enterprise, and track the status of the data source in real time to ensure the integrity and efficiency of data collection. For the real-time collected pharmaceutical sales data, the intelligent data cleaning module can be used to remove redundant, missing or inaccurate data to ensure the quality of the data.

[0035] In this exemplary embodiment, the cleaned data can be professionally processed and analyzed, using intelligent algorithms to identify potential problems or risks in the data and perform optimization processing.

[0036] In this exemplary embodiment, the pharmaceutical sales data is cleaned and standardized to obtain standardized data after removing redundant data, missing data, and inaccurate data, including:

[0037] Anomaly detection algorithms are used to identify outliers in the data or data that does not conform to regular patterns and remove them as inaccurate data.

[0038] For example, in pharmaceutical sales data, anomaly detection can identify unusual fluctuations, such as sudden increases or decreases in sales of a particular drug, allowing management to identify problems early and implement appropriate strategies. Available algorithms include Isolation Forest, Local Outlier Factor (LOF), and Z-score. These algorithms automatically identify unusual fluctuations in pharmaceutical sales, inventory, or other key metrics, providing timely alerts and preventing potential inventory overstocks or shortages.

[0039] In this exemplary embodiment, after data collection is completed, all data will enter the intelligent data cleaning and preprocessing module. This module first performs preliminary filtering on the collected raw data to remove redundant data and data with incorrect format. For example, duplicate records are deleted, and blank fields or irrelevant fields are removed. Secondly, the module will standardize the data to ensure that data from different data sources have a unified format, unit and identifier. At this time, the data cleaning module will apply data verification rules to check the integrity of the fields, the rationality of the data, and the detection of outliers. In this way, through the intelligent data cleaning and preprocessing mechanism, not only redundant and inaccurate data are automatically removed, but also standardized through the rule engine, automatically identifying and repairing anomalies or inconsistencies in the data, thereby providing a high-quality data foundation for subsequent data processing and analysis.

[0040] Among them, before generating a customized report, the system can perform a demand analysis based on the user's needs. The goal of this sub-step is to identify the user's specific needs through interaction with the user (for example, inputting the report type, required data dimensions, etc.) and selecting an appropriate report template. Report templates can be pre-designed in formats and structures based on different industry standards and user needs (such as sales reports, inventory reports, compliance reports, etc.). The system will select an appropriate template based on the needs and make necessary adjustments to it for subsequent data filling. For example, an automated demand identification algorithm is used to intelligently select and adjust report templates based on the business needs input by the user to improve the accuracy and customization of report generation.

[0041] Specifically, text classification algorithms (such as support vector machines or convolutional neural networks in deep learning) can be used to automatically understand the needs of user input. Through these algorithms, the system can analyze the natural language descriptions of user input (such as "I want a sales report", "I need to display inventory data by region and time dimensions", etc.) and map them to preset report template categories. For example, if a user requests "relationship analysis of sales and market investment by quarter", the system will identify "sales" and "market investment" as the main data dimensions and automatically select a suitable report template (such as a time series analysis chart or an association analysis report template). By using deep learning models, the system can more accurately understand the user's personalized needs and generate more customized report templates.

[0042] Furthermore, collaborative filtering algorithms can be used to predict the type of report template a user might need based on their historical needs and behaviors. By analyzing user preferences and behavior patterns, collaborative filtering technology recommends the report template that best meets their needs, thereby improving the accuracy and customization of report generation.

[0043] When analyzing user needs, cluster analysis can be used to group drug sales data based on characteristics such as region and sales type. For example, the system can use K-means cluster analysis to stratify sales data by region and drug type to identify high-performing and low-selling drug categories. Each cluster represents a specific sales pattern or market demand, helping users understand the distribution of business from a regional or drug category perspective.

[0044] At the same time, multiple regression analysis is used to analyze the relationship between multiple variables. In the scenario of drug sales, the system can use multiple regression analysis to identify the impact of different factors (such as market investment, seasonal changes, etc.) on sales. For example, through the regression model, the system can analyze the relationship between drug sales and multiple factors such as market investment, advertising expenditure, seasonal changes, etc., thereby generating a report that includes trend analysis and year-on-year and month-on-month changes. Through this method, the system can automatically generate forecast values for drug sales and provide data support for decision makers.

[0045] In this application, once the report template is determined, the system will then perform the process of data mapping and field matching. The goal of this sub-step is to match the cleaned data with the fields in the selected report template. The system automatically identifies the required fields in the template (such as drug name, sales quantity, market input, inventory, etc.) and maps the corresponding data source fields. This process ensures the consistency and accuracy of the data and the template. In this way, the need for manual operation is reduced through the automated data mapping and field matching mechanism, while the accuracy of data filling is improved, avoiding errors caused by manual operation.

[0046] After data mapping is complete, the system uses intelligent data analysis algorithms to perform multi-dimensional analysis of the data. For example, the system can generate reports based on drug sales data, displaying tiered data by region, time period, drug type, and other dimensions, helping users understand business performance from different perspectives. The analysis results are automatically populated into the corresponding report template. Report content includes not only numerical data but also trend analysis, year-on-year and month-on-month changes, and other information. In this way, by utilizing multi-dimensional intelligent data analysis technology, multi-angle data can be automatically analyzed and presented, improving decision-making accuracy.

[0047] In some embodiments, obtaining pharmaceutical sales data from multiple pharmaceutical industry data sources includes:

[0048] Monitor data source status in real time and dynamically adjust the collection priorities of multiple pharmaceutical industry data sources;

[0049] Pharmaceutical sales data based on the multiple pharmaceutical industry data sources.

[0050] In this exemplary embodiment, when performing real-time data collection, raw data for the pharmaceutical industry can be obtained from various selected data sources through scheduled tasks or real-time data stream interfaces. This data may include sales data, inventory data, drug information, market trends, etc. The system obtains data in real time through API interfaces, data crawling, or docking with partner data systems. The data collection module supports parallel collection from multiple channels to ensure the timeliness and comprehensiveness of data. In this way, through the real-time data collection mechanism, not only can data be automatically acquired from multiple data sources, but it can also process large-scale, high-frequency data streams to ensure the timeliness and accuracy of data without causing system overload or data loss.

[0051] In some embodiments, the pharmaceutical sales data is cleaned and standardized to obtain standardized data after removing redundant data, missing data, and inaccurate data, including:

[0052] Use the rule engine to detect and repair pharmaceutical sales data with missing fields, format errors, and abnormal values;

[0053] Through the standardization module, the data format of the repaired pharmaceutical sales data is unified and the timestamps are aligned to obtain standardized data with time consistency.

[0054] In this exemplary embodiment, the repairing of pharmaceutical sales data with missing fields includes:

[0055] If the data type of the missing field is a key field, the pharmaceutical sales data of the missing field is filled based on predefined rules; wherein the predefined rules include at least one of the following: filling the missing value based on the mean of similar records, forward filling, and backward filling;

[0056] If the data type of the missing field is a non-key field, the pharmaceutical sales data with the missing field will be marked and reported.

[0057] In this exemplary embodiment, during the data standardization process, the data can be checked for missing values, particularly missing key fields (e.g., drug number, sales amount, etc.). If missing values are found, they can be filled according to predefined rules or corresponding algorithms. For example, missing values can be filled based on the mean of similar records, or by using forward or backward filling methods. Furthermore, missing values that cannot be effectively filled are marked as "missing data" by the system for subsequent processing or manual intervention. In this way, through an intelligent missing value handling mechanism, different filling strategies are adopted based on data type and business needs, preventing missing data from significantly impacting analysis results. Specifically, the system may perform missing value analysis on the processed data to determine whether key fields (e.g., drug number, sales amount, inventory level, etc.) are missing. If missing values are in non-key fields, the system will flag and report them; if missing values are in key fields, the system will implement intelligent filling strategies. Filling methods can include data extrapolation (e.g., estimating missing values using data from adjacent time periods), filling with the mean or median, or using algorithms (e.g., regression analysis or machine learning) for predictive filling. In this way, by introducing intelligent algorithms, different missing value processing strategies are adopted for different data scenarios to ensure that the data is as complete as possible without introducing obvious errors.

[0058] In this exemplary embodiment, repairing the pharmaceutical sales data with abnormal values includes performing data repair based on any of the following methods:

[0059] Replace the outliers according to the mean of similar data;

[0060] Directly mark it as an exception and conduct manual review and modification;

[0061] Make reasonable estimates and repairs after comparing with historical data;

[0062] Replace data values with inferred values based on data analysis.

[0063] In this exemplary embodiment, after data standardization and missing value processing, the system enters the outlier detection phase. This phase automatically identifies outliers in the data using pre-defined rules or machine learning algorithms. The system automatically verifies and corrects detected outliers based on a rule base. Correction methods include marking them as abnormal records, performing reasonable corrections after comparing them with historical data, or replacing data values with estimated values based on data analysis. In this way, machine learning algorithms and rule engines intelligently detect and correct outliers, minimizing human error and data inconsistencies and ensuring data reliability. For example, if sales volume or amount exceeds the normal range, or if data such as drug prices and inventory deviate significantly from historical trends, the system will flag them as outliers. For detected outliers, the system can take corrective measures, such as replacing them based on the mean of similar data, or directly marking them as abnormal and conducting manual review. In this way, by integrating rule-based and machine learning outlier detection mechanisms, the system can automatically identify and correct potential anomalies in the data, reducing manual intervention and improving data reliability.

[0064] After data cleansing, the system enters the standardization phase. This standardization module unifies data formats from different sources, ensuring that all data fields, units, codes, and timestamps conform to industry or internal company standards. For example, fields such as drug name, strength, and batch number in sales data are standardized to a unified naming convention to ensure data consistency. Furthermore, time fields are standardized to ensure consistent timestamps across different data sources, facilitating subsequent time series analysis. This intelligent data standardization solution, through standardized formats, units, and timestamps, ensures seamless integration of multi-channel and multi-source data, facilitating subsequent analysis and report generation. Specific operations may include categorizing data by type and dimension, unifying numerical types (e.g., amount, time, quantity), and structuring the data into formats suitable for subsequent analysis (e.g., two-dimensional tables, time series, matrices, etc.). This step ensures that data from different sources and formats can be presented uniformly and neatly. Through intelligent data preprocessing mechanisms, the system can flexibly select formatting methods based on the characteristics of the data and analytical requirements, achieving automatic data normalization.

[0065] After cleaning, standardization, missing value processing, and outlier correction, the data is stored in a dedicated database. At this point, the data is accurate, consistent, and compliant with standards, providing a reliable foundation for subsequent data processing, analysis, and report generation. The system categorizes and stores the cleaned data according to different data types, ensuring that the data storage structure supports fast retrieval and subsequent data processing needs. This efficient database storage structure ensures rapid access to cleaned data and provides efficient data access for subsequent data analysis and report generation.

[0066] In some embodiments, generating structured data by performing multi-dimensional analysis and feature extraction on the standardized data includes:

[0067] performing a correlation analysis between different data fields in the standardized data based on a Pearson correlation coefficient, a Spearman rank correlation coefficient, and / or a point-biserial correlation coefficient to obtain correlation analysis results between the different data fields;

[0068] By performing principal component analysis on the correlation analysis results between different data fields in the standardized data, characteristic fields are extracted where the correlation degree between different data fields in the standardized data exceeds a predetermined value;

[0069] The structured data is generated based on a characteristic field in which the degree of correlation between different data fields in the standardized data exceeds a predetermined value.

[0070] In this exemplary embodiment, when performing multi-dimensional analysis and feature extraction on the standardized data, natural language processing can also be used to analyze and process text data to extract valuable information from the data. For example, natural language processing can be used to analyze text information such as customer reviews and drug instructions to extract key content such as drug characteristics or side effects that users are concerned about. Bag of Words, TF-IDF, and Word2Vec are available NLP (Natural Language Processing) algorithms. These algorithms help pharmacies analyze customer needs, identify market trends, and further optimize products and services.

[0071] In this exemplary embodiment, data correlation analysis is used to discover the intrinsic connections between different data items. The system identifies highly correlated fields (such as drug sales and market investment, drug sales and seasonal changes, etc.) by calculating the correlation coefficients between each field. Then, based on the analysis results, the system automatically extracts relevant features and provides key variables for subsequent decision support and report generation. This process provides the basis for subsequent multi-dimensional report generation and data analysis. In this way, by using intelligent algorithms to automatically calculate the correlation between data and extract characteristic variables that are critical to business decisions, deep data mining and optimization are achieved.

[0072] Highly correlated fields identified through data correlation analysis (e.g., the relationship between drug sales, marketing investment, and seasonal variations) are directly incorporated as core variables into the report generation framework. These variables are no longer presented as isolated data, but rather as correlated relationships. For example, report charts (such as scatter plots and heat maps) can visualize the strength of correlation between variables, or automatically generate conclusions such as "for every 10% increase in marketing investment, a particular drug's sales increase by 8%" within analysis paragraphs. Furthermore, key features (e.g., key factors influencing sales) identified during feature extraction using techniques such as PCA (Principal Component Analysis) and mutual information analysis are prioritized and integrated into the report's multidimensional analysis module. For example, in sales reports, the system automatically layers data by "highly correlated features" (e.g., season, promotional activity), allowing users to quickly identify key drivers influencing business operations. This correlation allows reports to not only present data results but also reveal the business logic behind the data, enhancing the report's support for decision-making.

[0073] The results of correlation analysis and feature extraction are not only used to demonstrate associations but also serve as input data to support subsequent forecasting models (such as multivariate regression and machine learning algorithms), which then output forecast conclusions and action recommendations through reports. For example, by identifying "strong correlations between sales, historical data, and market investment," the system can build a forecasting model based on historical data, generating a forecast result in a report such as "sales of a certain drug are expected to increase by 15% next quarter," and annotating key influencing factors (such as the driving effect of promotional activities).

[0074] If feature extraction finds that "inventory turnover rate is highly correlated with the expiration date of drugs and the supplier's delivery cycle", the report can automatically generate procurement recommendations (for example, it is recommended to purchase drugs with a shorter expiration date three months in advance to avoid inventory backlogs), or mark the "inadequate inventory risk" through the risk assessment module and link it to the procurement system.

[0075] When the correlation analysis shows that "the sales volume of a certain type of drug is strongly correlated with the regional population density and the degree of aging", the report can recommend "increasing the distribution volume of this drug in areas where the aging rate exceeds 20%" based on the characteristic data.

[0076] These predictions and suggestions are not independent functions. Instead, they are presented in the form of visual charts, smart prompts or conclusion paragraphs through the deep integration of the report generation module and data analysis results, upgrading the report from a data aggregation tool to an intelligent decision-making engine.

[0077] The goal of data correlation analysis is to identify variables with significant correlations by calculating correlation coefficients between data fields. In pharmacy operations, analyzing the relationship between drug sales and factors such as market input and sales seasonality helps understand the impact of different variables on drug sales.

[0078] Through the correlation analysis method, the system can automatically calculate the correlation between fields and identify highly correlated field pairs, providing a basis for subsequent feature extraction and report generation.

[0079] Among them, the purpose of feature extraction is to identify features with significant influence from a large amount of data, which are crucial for decision support. Based on the results of data correlation analysis, the system will select those data fields that are highly correlated and can effectively reflect business behavior as features. Feature extraction methods include principal component analysis, which maps high-dimensional data to low-dimensional space while retaining as much data variance as possible. PCA selects the most representative features by calculating the eigenvalues and eigenvectors of the data covariance matrix. For example, through PCA, the system can extract the most explanatory features, making subsequent analysis more efficient. In addition, mutual information is another algorithm for evaluating the dependency between two variables. It can not only measure linear relationships, but also reveal nonlinear relationships. The formula for mutual information is as follows:

[0080] Here, X and Y are two random variables (for example, in drug sales analysis, these could be the sales volume of a drug and the purchasing behavior of customers), P(x,y) is the joint probability, which represents the probability of the two random variables occurring simultaneously, and P(x) and P(y) are the marginal probability distributions of X and Y, respectively, representing the probability of the random variables X = x and Y = y occurring individually.

[0081] Mutual information is particularly effective in analyzing drug sales data and customer behavior, and is helpful in identifying important features that affect drug sales.

[0082] Another feature extraction method is Lasso regression (Least Absolute Shrinkage and Selection Operator, Lasso), which uses L1 regularization to select features and reduce redundant features. The principle of Lasso regression is to impose a penalty on the regression coefficients, forcing some feature coefficients to zero, thereby achieving feature screening. It is particularly suitable for high-dimensional data.

[0083] Based on correlation analysis and feature extraction, the system needs to further process the selected features to provide structured data support for subsequent model training and report generation. Feature engineering methods include standardization and normalization, which are used to convert features of different dimensions to the same scale to ensure that the weights of different features are consistent. Standardization usually uses the Z-score normalization method to convert the data into a distribution with a mean of 0 and a standard deviation of 1. Normalization usually uses Min-Max normalization to scale the data to a range of 0 to 1. In addition, category encoding is a method of converting categorical variables into numerical features. Encoding methods include One-Hot Encoding and Label Encoding, which can convert categorical variables into a form acceptable to machine learning models.

[0084] For example, the different data fields in the standardized data include at least a drug name field, a drug type field, a drug applicable population field, a drug applicable symptom field, a sales quantity field, a market input field, a drug sales seasonality field, and an inventory field;

[0085] The correlation analysis between different data fields in the standardized data is performed based on the Pearson correlation coefficient, the Spearman rank correlation coefficient and / or the point-biserial correlation coefficient to obtain the correlation analysis results between different data fields, including:

[0086] Based on the Pearson correlation coefficient, Spearman rank correlation coefficient and / or point-biserial correlation coefficient, a correlation analysis is performed among the drug name field, drug type field, drug applicable population field, drug applicable symptom field, sales quantity field, market input field, drug sales seasonality field, and inventory field in the standardized data to obtain a correlation analysis result among the drug name field, drug type field, drug applicable population field, drug applicable symptom field, sales quantity field, market input field, drug sales seasonality field, and inventory field in the standardized data;

[0087] The extracting of characteristic fields whose correlation degree between different data fields in the standardized data exceeds a predetermined value by performing principal component analysis on the correlation analysis results between different data fields in the standardized data comprises:

[0088] If principal component analysis is performed on the correlation analysis results between different data fields in the standardized data, and it is determined that the degree of correlation between the drug sales seasonality field and more than half of the drug name field, drug type field, drug applicable population field, drug applicable symptom field, sales quantity field, market input field and inventory field exceeds a predetermined value, then the drug sales seasonality field is extracted.

[0089] In some embodiments, performing principal component analysis on the correlation analysis results between different data fields in the standardized data to extract characteristic fields in which the correlation degree between different data fields in the standardized data exceeds a predetermined value includes:

[0090] If principal component analysis is performed on the correlation analysis results between different data fields in the standardized data, and it is determined that the degree of correlation between the target field and more than half of the different data fields in the standardized data exceeds a predetermined value, the target field is extracted; wherein the target field is one of the data fields in the standardized data.

[0091] In this exemplary embodiment, the structured data is matched with the template fields in the user demand report template, the structured data is mapped to the user demand report template, and the compliance check and risk assessment rules embedded in the user demand report template are combined to generate a personalized report, including:

[0092] If there is field data in the structured data that does not comply with the compliance check and risk assessment rules, then the drugs associated with the fields that do not comply with the compliance check and risk assessment rules are marked as potential risks;

[0093] The structured data mapping is performed based on the fields marked with potential risks to generate a personalized report with potential risk prompts.

[0094] In this exemplary embodiment, the system uses rule-based or machine learning-based risk assessment models to identify potential risks within the data. For example, the system can use historical data and market trends to determine whether sales of certain drugs are lower than expected, or whether marketing investment in certain drugs is excessive, resulting in wasted resources. These potential risks are flagged and correlated with other business indicators to generate personalized reports with potential risk indicators. This process supports subsequent compliance checks, resource allocation, and strategic decision-making.

[0095] The system identifies potential risks by analyzing historical data and market trends. These risks include issues such as substandard drug sales and excessive market investment. The resulting risk prediction not only identifies risky drugs but also has a profound impact on subsequent report generation. Specifically, when generating customized reports, the system embeds this risk information into the report content. For example, the system may add additional risk markers to the sales data of certain drugs, alerting decision makers to the potential risk of declining sales or wasted resources. The report generation module automatically presents this risk information to users and provides detailed risk analysis charts in the report, such as risk trend graphs and lists of risky drugs. This not only allows decision makers to clearly understand potential issues but also provides a basis for further resource adjustments and procurement decisions. Therefore, risk identification is not just a standalone step; it directly influences the content of the report and provides decision makers with real-time risk assessment support.

[0096] During the report generation process, risk assessment functionality is embedded through intelligent mechanisms. Specifically, the system analyzes and compares data based on pre-set compliance rules or risk patterns learned through machine learning models. These rules can include expected standards for drug sales, a reasonable range for market investment, and expectations for seasonal fluctuations. During this process, the system compares actual data with these standard data, for example, comparing the difference between actual and expected drug sales or whether market investment exceeds the budget. If a drug's sales are lower than expected, or if market investment is far above the reasonable range, the system will mark these drugs as potential risks. Risk assessment results are automatically highlighted in the generated reports, allowing managers to prioritize these risks in subsequent decision-making.

[0097] During specific operations, the system embeds compliance rules into report templates. For example, in a sales report, the system automatically identifies drugs that fail to meet standards or have excessively high market investment, based on sales volume and market investment, and displays corresponding risk warnings in the report. This way, reports not only provide general business data but also provide real-time risk warnings.

[0098] Furthermore, the integration of risk identification with drug procurement and sales optimization primarily impacts procurement planning, sales strategies, and inventory management. By predicting potential risks, the system can assist decision-makers in timely adjusting procurement and sales strategies. For example, if the system identifies certain drugs with substandard sales and excessive marketing investment, these drugs will be marked as risky. Based on this information, decision-makers can make appropriate procurement adjustments, reducing inventory overstocks or halting further purchases. Furthermore, the sales department can adjust promotional strategies or optimize marketing investment based on the risk assessment results to improve drug sales performance.

[0099] Furthermore, the system can automatically optimize procurement and sales through integration with supply chain management systems. For example, based on risk assessment reports, the system can recommend reducing the purchase volume of certain drugs or increasing promotional efforts. This ensures sufficient inventory while avoiding the waste of funds or unsold inventory caused by over-purchasing. In this way, risk assessment not only influences report generation but also directly participates in optimizing drug procurement and sales strategies, providing comprehensive decision-making support for enterprises.

[0100] Compliance checks and risk assessments are automatically embedded in the report generation process to ensure that all generated reports adhere to industry standards and policy requirements. This process not only helps ensure report compliance but also identifies potential risks and provides decision support.

[0101] Before generating a report, the system will first identify the compliance standards that need to be met in the report based on preset industry policies and legal and regulatory requirements. Specifically, the system will map industry regulations (such as the Drug Administration Law, sales compliance regulations, price regulations, etc.) and internal corporate compliance requirements to the fields, formats, and data standards required in the report. The system automatically compares the content of the generated report with these compliance requirements to check whether there are any violations of policy regulations. In this way, through the automated compliance requirement identification mechanism, industry standards and policy requirements are directly integrated into the report generation process, achieving full automation of compliance checks and avoiding omissions or errors in manual operations.

[0102] After the compliance check is completed, the system proceeds to identify and assess risks in the generated reports. This process involves analyzing the data in the reports through intelligent algorithms to identify content that may pose business risks. For example, the system will evaluate whether drug sales are consistent with market trends, analyze whether procurement costs exceed the budget, and check for unreasonable market investment. The system can also identify potential market and operational risks based on historical data and current data patterns, and mark them in the reports for reference by relevant decision makers. In this way, through the intelligent risk identification and assessment module, the system can automatically analyze potential risks in the data and conduct risk prediction based on historical trends and market dynamics, thereby improving the company's early warning and decision-making support capabilities in operations.

[0103] After identifying potential risks, the system will provide specific corrective suggestions or optimization measures. For example, if the procurement cost of a particular drug is found to be too high, the system will recommend optimizing the procurement process or adjusting the pricing strategy. If the return on investment in certain markets is found to be low, the system will recommend adjusting the marketing strategy. In addition, based on the assessment results, the system will automatically generate a detailed risk assessment report, including a detailed description of each risk, its causes, possible consequences, and corrective measures. In this way, the system not only identifies risks but also automatically generates optimization suggestions based on data analysis, providing companies with specific operational guidance, allowing them to adjust their strategies in real time to mitigate potential risks.

[0104] After completing all compliance and risk assessments for this application, the system will generate a compliance and risk assessment report. This report will detail the results of all compliance checks, the identified risks and their levels, and any relevant remediation recommendations. This report will be automatically embedded in the final customized report, ensuring that it not only displays data but also provides important information on legal compliance and risk management. This automated generation of a comprehensive report encompassing compliance and risk assessments reduces the time and cost of manual review while ensuring the compliance and quality of the report.

[0105] Through the compliance and risk assessment process of this invention, enterprises can automatically perform compliance checks on their reports and promptly identify and assess potential risks. Through intelligent risk assessment and compliance matching, the system ensures that all reports not only comply with industry standards and policy requirements, but also provides enterprises with early warnings and optimization recommendations for potential risks. This process reduces manual intervention, improves the accuracy and reliability of reports, and provides enterprises with powerful decision-making support and compliance assurance, helping them more effectively manage risks and adjust strategies in a complex and volatile market environment.

[0106] After data processing and analysis are complete, the system generates analysis results and outputs them to the report generation module. These data processing results include cleaned data, supplemented and revised datasets, correlation analysis results, feature extraction, and risk assessments. The system automatically generates optimized datasets based on reporting requirements and supports customized report generation in the next phase. The system also provides data visualizations, enabling users to quickly understand the data analysis results. Through optimized data output capabilities, the system not only provides standardized processing results but also supports the generation of data visualization charts, enabling users to intuitively understand the data and analysis results.

[0107] In some embodiments, performing correlation analysis between different data fields in the standardized data based on the Pearson correlation coefficient, the Spearman rank correlation coefficient, and / or the point-biserial correlation coefficient to obtain correlation analysis results between different data fields includes:

[0108] Determine the linear correlation between continuous variables in the standardized data using the Pearson correlation coefficient to obtain a first correlation analysis result between data fields;

[0109] Determining the monotonic relationship between the non-normally distributed data in the standardized data by using the Spearman rank correlation coefficient to obtain a second correlation analysis result between the data fields;

[0110] The correlation between continuous variables and binary variables was determined by point-biserial correlation coefficient, and the third correlation analysis results between data fields were obtained;

[0111] Based on the first correlation analysis result, the second correlation analysis result and the third correlation analysis result, a correlation analysis result between the different data fields is obtained.

[0112] In this exemplary embodiment, the correlation is calculated by measuring the linear relationship between two variables using the Pearson correlation coefficient, which is applicable to continuous variables. The value of the Pearson coefficient is between -1 and 1, indicating a linear correlation between the variables.

[0113] In addition, the Spearman rank correlation coefficient is a measure suitable for nonlinear relationships, especially when the data does not conform to the normal distribution, revealing the monotonic relationship between variables. Its calculation formula is as follows:

[0114]

[0115] Here di is the rank difference of each pair of variables, n is the sample size, and ρ is the Spearman rank correlation coefficient, which ranges from -1 to 1.

[0116] Finally, the point-biserial correlation coefficient was used to analyze the correlation between a continuous variable and a binary variable, such as the relationship between whether a drug was on promotion and its sales volume.

[0117] In this exemplary embodiment, the correlation analysis results between different data fields may include any one or any combination of a first correlation analysis result, a second correlation analysis result, and a third correlation analysis result, wherein the first correlation analysis result, the second correlation analysis result, and the third correlation analysis result represent the correlation relationship between different types of data in the standardized data.

[0118] In some embodiments, before generating a personalized report based on the compliance check and risk assessment rules embedded in the user demand report template, the method includes:

[0119] By using an autoencoder to learn a low-dimensional representation of the data, field data in the structured data that does not comply with compliance inspection and risk assessment rules is identified; wherein the compliance inspection and risk assessment rules include at least one of the following: industry regulations, internal corporate regulations, a reasonable range of market investment, an expected range of seasonal fluctuations, and an expected standard for drug sales.

[0120] In this exemplary embodiment, autoencoders are a type of unsupervised deep learning algorithm that can be used for data dimensionality reduction and anomaly detection. When processing pharmaceutical sales data, pharmacies face a large amount of multidimensional data, including sales figures, inventory, marketing inputs, and other factors. Autoencoders can identify unusual patterns or potential risks in the data by learning a low-dimensional representation of the data.

[0121] For example, using autoencoders, the system can extract the most representative features from large amounts of pharmaceutical sales data, reducing dimensionality while retaining key information. This allows deep learning models to process and analyze data more efficiently, avoiding interference from redundant information. Furthermore, autoencoders can identify unusual fluctuations or unusual patterns in sales data, such as sudden spikes or drops in sales of certain drugs. These anomalies allow managers to promptly identify potential market issues or inventory overhang risks. During this process, the system compares actual data with this benchmark data, for example, comparing actual sales against projected sales or whether marketing investment exceeds budget. If a drug's sales fall short of expectations or marketing investment far exceeds reasonable limits, the system identifies it as a potential risk. Risk assessment results are automatically highlighted in generated reports, allowing managers to prioritize these risks in subsequent decision-making.

[0122] In some embodiments, the method comprises:

[0123] When generating personalized reports, a sales trend analysis is performed on the drug sales time series data in the standardized data based on a convolutional neural network and / or a recurrent neural network to extract the periodic variation characteristics of sales peaks and valleys, and the sales trend of the drug within a predetermined future time is predicted based on the periodic variation characteristics of sales peaks and valleys;

[0124] The generation of a personalized report based on the compliance check and risk assessment rules embedded in the user demand report template includes:

[0125] Based on the compliance check and risk assessment rules embedded in the user demand report template, the forecast results of the sales trend of the drug within a predetermined time in the future are added to generate a personalized report.

[0126] In this exemplary embodiment, the CNN (Convolutional Neural Network) is primarily used for image recognition and processing, but it also demonstrates strong capabilities in processing one-dimensional data, such as trend analysis and chart pattern recognition of pharmaceutical sales data. In pharmaceutical sales data, the CNN can be applied to sales trend analysis, learning how pharmaceutical sales change over time and helping the system identify patterns in sales fluctuations. This is not limited to processing time series data; it can also identify sales details in different dimensions (e.g., region, season, etc.), providing precise guidance for inventory management and promotional strategies.

[0127] For example, CNNs can identify seasonal fluctuations in pharmaceutical sales. By converting sales data from different time periods into an image-like matrix, CNNs can extract characteristics of sales peaks and valleys and cyclical changes, thereby predicting future sales trends. This is crucial for pharmacies to make proactive purchasing decisions and inventory adjustments.

[0128] Furthermore, RNNs (Recurrent Neural Networks), particularly LSTMs (Long Short-Term Memory), are well-suited for processing time-dependent sequential data. Pharmaceutical sales data typically exhibits distinct time series characteristics, with sales volumes influenced by factors such as seasonality, marketing campaigns, and holidays. Therefore, RNNs are particularly well-suited for predicting future pharmaceutical sales.

[0129] By memorizing past data points, RNNs can incorporate historical information when processing current data. LSTM networks are particularly adept at capturing long-term temporal dependencies, providing high accuracy for drug sales forecasts. For example, pharmacies can use LSTM to predict sales of a particular drug over the next few months, enabling them to make precise inventory management and procurement plans based on historical sales data and other external factors (such as advertising and holiday promotions).

[0130] Deep learning algorithms can also be tightly integrated with report generation modules to enable automated and intelligent report customization. During report generation, deep learning models can predict drug sales trends based on historical data and provide data support for reports. These predictions can be directly embedded in generated reports, providing decision makers with real-time market dynamics and trend analysis, helping them make more accurate purchasing, inventory, and marketing decisions.

[0131] Through deep learning, reports can not only display traditional sales data but also automatically generate trend forecasts, anomaly detection, and potential risk analysis. For example, based on the predictions of the deep learning model, the system can automatically add sales forecasts for the next few months, market return on investment, and other information to the report, making the report content more comprehensive and accurate.

[0132] Deep learning can also automate compliance checks. Compliance is crucial in the pharmaceutical industry, as any failure to adhere to policies can result in legal risks or fines. Deep learning algorithms can analyze sales data and marketing investments to see if they meet regulatory requirements and automatically generate compliance reports. For example, using deep learning, the system can analyze the relationship between a drug's marketing investment and sales performance to determine whether marketing investment is excessive or unreasonable, thereby preventing waste of corporate resources or policy violations.

[0133] By combining deep learning with compliance detection modules, the system not only detects non-compliant data but also generates compliance reports that meet industry standards, ensuring their compliance and effectiveness. This not only improves data processing efficiency but also reduces the complexity and error rate of manual operations.

[0134] Deep learning models can also serve as intelligent decision-making support tools, helping pharmacy management make more accurate strategic decisions. By analyzing historical data, customer behavior, and market trends, deep learning algorithms can provide management with data-driven decision-making recommendations. For example, deep learning can predict the sales potential of a particular medication, thereby supporting purchasing decisions and promotional strategies. Furthermore, deep learning can identify slow-moving or over-purchased items, providing timely adjustment recommendations and avoiding wasted resources.

[0135] By integrating deep learning models, the present invention can not only improve the intelligence level of report customization, but also provide pharmacies and merchants with strategic decision-making support based on big data and deep learning, thereby improving operational efficiency and competitiveness.

[0136] In short, the specific application of deep learning algorithms in pharmaceutical sales data processing can improve the accuracy of data analysis, discover potential sales patterns, predict future sales trends, identify abnormal patterns in data, and provide intelligent solutions for report generation, compliance checking, and decision support.

[0137] In this exemplary embodiment, generating a personalized report includes:

[0138] After mapping the structured data to the user demand report template, determining the report output format when generating the report;

[0139] After the report output format is determined, a visual chart related to the report data is generated; wherein the chart includes at least one of the following: a bar chart, a pie chart, and a line chart;

[0140] The chart is embedded in the report to support the visual presentation of data.

[0141] In this exemplary embodiment, Figure 3 This is a flow chart of report output and optimization according to an embodiment of the present invention. Figure 3 As shown, the report output and optimization process includes:

[0142] Step 30: Select the report output format;

[0143] Step 31: Data visualization and chart generation;

[0144] Step 32: Report content optimization and data analysis suggestions;

[0145] Step 33: Report compliance check and review;

[0146] Step 34: Report distribution and sharing.

[0147] Once the report output format is determined, the system automatically generates visualization charts related to the report data. These charts, including bar charts, pie charts, and line charts, help users intuitively understand data distribution, trends, and key indicators. These charts are automatically embedded in the report to support data visualization. Through intelligent chart generation and automatic embedding, the system presents complex data through intuitive charts, making reports easier to understand. Especially when multi-dimensional data is involved, the use of charts greatly enhances the intuitiveness of decision support.

[0148] During the report generation process, users can further adjust the report's style and format to meet their specific presentation needs. For example, they can adjust fonts, colors, table borders, and chart styles to ensure that the report meets the user's or company's internal visual standards. These style adjustments are automated based on templates, providing users with a more personalized reporting experience without changing the core content of the report. Through intelligent style optimization and customization, the system not only enables users to obtain functional reports but also meets their personalized visual needs, improving the user experience and report readability.

[0149] The system automatically populates the report with processed and analyzed data based on matched fields. During the automatic filling process, the system performs data validation to ensure that the filled content conforms to the report template format and that data integrity is guaranteed. If any inconsistencies or errors are detected in the filled data, the system automatically prompts and makes corrections or annotations. This helps ensure that the generated report content is complete and accurate, avoids errors or omissions, and reduces the need for manual intervention.

[0150] After all data is populated and formatted, the system generates the final report and provides various export and distribution options. Reports can be exported to formats such as PDF, Excel, and CSV, allowing users to archive and share them based on their needs. The system also supports automatically sending reports to designated email addresses or distributing them to relevant personnel through other channels.

[0151] Through the customized report generation process of the present invention, the system can automatically select report templates, perform data mapping, analyze data, and generate multi-dimensional reports based on user needs. Furthermore, through intelligent report style optimization and customization, reports are not only highly personalized but also ensure they meet the visual and business requirements of different users. Automatic fill and data validation functions ensure the accuracy and reliability of reports, reduce manual intervention, and improve the usability and dissemination efficiency of reports through multiple export and distribution methods. This series of innovative steps not only improves the efficiency of report customization, but also effectively reduces the error rate and enhances data analysis and decision support capabilities, thereby optimizing the management and operational processes of pharmaceutical companies.

[0152] After report customization is complete, the output format can be determined based on user needs. The system supports multiple output formats, such as PDF, Excel, CSV, and Word, allowing users to select the appropriate format for export based on their needs. Furthermore, the system provides specific output templates to ensure that report formats are consistent and meet internal company requirements. By providing multiple export formats, the system can adapt to the needs of different users and ensure that reports are suitable for various usage scenarios, such as internal reporting, external audits, or sharing with partners.

[0153] In this exemplary embodiment, generating a personalized report includes:

[0154] After generating the personalized report, by collecting feedback data from users after using the report, using machine learning and intelligent algorithms, the report structure, content and presentation method are optimized in real time; wherein, the feedback data from users after using the report includes at least one of the following: report usage frequency, modification history, and user preferences.

[0155] After generating reports and charts, the system optimizes the content to ensure a more readable and professional presentation. Data fields, charts, and text descriptions within the reports are adjusted to meet user needs, making them more relevant to actual applications. Furthermore, the system provides optimization suggestions based on data analysis, enabling users to make more effective decisions based on the reports. For example, the system may provide recommendations for optimizing inventory or adjusting marketing strategies based on sales data.

[0156] Before report output, the system performs another compliance check to ensure that the generated report complies with relevant laws, policies, and internal company standards. This step primarily involves automated verification to ensure that the report content does not violate compliance requirements, such as those related to pharmaceutical sales and inventory management. Simultaneously, the system generates a compliance review report for review and confirmation by relevant personnel. This automated integration of compliance checks provides assurance that reports are compliant, reducing the time and workload of manual review and ensuring that report content adheres to industry regulations.

[0157] After report generation and optimization is complete, the system automatically supports report distribution and sharing. Users can choose to send reports directly to designated individuals or share them with relevant departments, partners, or auditors via email, cloud platforms, and other methods. The system also supports permission control to ensure that only authorized personnel can access and view reports. This automated distribution and sharing streamlines the report delivery process while ensuring data security and privacy, thereby improving work efficiency.

[0158] Through the report output and optimization process of the present invention, the system can not only ensure that the report is generated in a high-quality format, but also automatically generate intuitive data charts, allowing users to better understand complex data and improve decision-making efficiency. When optimizing the report content, the system will also provide optimization suggestions based on data analysis, which enhances the practicality and guidance of the report. The automated embedding of compliance checks enables reports to seamlessly meet industry standards and policy requirements, reducing the workload of manual intervention and review. Finally, the automated report distribution and sharing functions improve work efficiency, ensure that relevant personnel can obtain and use reports in a timely manner, and promote information flow and collaboration among teams. All these innovative features not only improve the efficiency and quality of report generation, but also simplify the company's report management process.

[0159] Through a dynamic feedback mechanism, the generated reports are continuously self-optimized and adjusted. Specifically, the system collects feedback data from users after using the reports (such as report usage frequency, modification records, user preferences, etc.), and uses machine learning and intelligent algorithms to optimize the report structure, content and presentation in real time. For example, the system can automatically adjust the presentation of report content (such as switching charts, reordering data, highlighting key data, etc.) based on historical data trends and user feedback, and provide improvement suggestions to users. In this way, by introducing intelligent feedback and self-optimization mechanisms, reports can be continuously adjusted as use deepens and user needs change, enhancing the intelligence and personalization of reports, improving user experience and enhancing the adaptability of decision-making.

[0160] In this application, a dynamic feedback mechanism is used to continuously optimize and adjust the generated reports. The system collects feedback data from users after using the reports (such as report usage frequency, modification history, user preferences, etc.), and uses machine learning and intelligent algorithms to optimize the report structure, content, and presentation in real time, thereby enhancing the intelligence and personalization of the reports, improving the user experience, and enhancing the adaptability of decision-making.

[0161] First, all data on user interactions with the report is collected, including the frequency of report use, user modification records in the report, preferences for data presentation (for example, whether a bar chart is preferred over a line chart), and specific adjustments made by the user. The system automatically records and stores this feedback data based on each user's behavioral characteristics and analyzes it. Through machine learning algorithms, the system can identify trends and preferences from user feedback to understand user needs. In this way, through machine learning and big data analysis, the system can identify users' personalized needs, capture patterns in user behavior, and ensure that reports can be gradually adapted and optimized to meet users' actual usage needs.

[0162] Based on user feedback data and historical data trends, the system will adjust the content of the report in real time and optimize its presentation. For example, if a user frequently views a data field and makes modifications, the system can automatically advance the display position of the field to ensure that it is more prominent in the report. If the display method of certain charts is frequently modified by users, the system will also adjust the default chart type according to the user's preferences (for example, converting the original bar chart to a pie chart or line chart) to make it more in line with the user's viewing habits. In this way, by constantly learning from the user's adjustment behavior and preferences, the report content is adaptively optimized to ensure that the report display meets personalized needs, and by adjusting field positions, chart types, and other means to improve data visibility and report readability.

[0163] Based on user feedback and behavioral analysis, the system automatically adjusts the chart type, data arrangement order, etc. in the report. For example, when the system finds that a certain data field has a high level of user attention, the system will automatically convert the field into a chart display or display it in a highlighted manner. At the same time, the type of chart will be automatically selected according to user preferences (for example, if the user prefers a bar chart, the system will automatically set the chart format to a bar chart). In addition, the system will automatically optimize the order of data display based on the nature of the data to ensure that key information is more prominent. In this way, by adjusting the chart type and data display method in real time, the interactivity and dynamism of the report are enhanced, automatically adapting to user preferences, and improving the intuitiveness and effectiveness of data display.

[0164] Based on user usage habits and data trend analysis, the system will regularly generate personalized report improvement suggestions. Users can further adjust the report content and presentation based on these suggestions. For example, the system may suggest combining certain data fields or optimizing the font style or color within the report to make the report clearer and more concise. Furthermore, the system will dynamically optimize the report structure based on user feedback, automatically identifying and removing redundant information to improve report focus and information delivery efficiency.

[0165] After optimization is complete, the system engages in continuous self-learning to adapt to the changing needs of more users. Each time a user uses a report and provides feedback, the system updates its learning model based on this new feedback, continuously improving its report generation and optimization capabilities. Through self-learning and adaptation, the system automatically adjusts its optimization strategies to ensure that reports remain synchronized with user needs and environmental changes. In this way, through continuous self-learning and adaptive adjustments, the system can continuously optimize its report customization strategies in different usage scenarios and provide efficient support in an ever-changing business environment, ensuring that reports remain efficient and adaptable over the long term.

[0166] Through the dynamic intelligent feedback and report self-optimization process of the present invention, the system can continuously adjust and optimize the report content according to user behavior and feedback, so that it always adapts to the actual needs of the user. This not only improves the personalization and intelligence of the report, but also allows the report to be gradually optimized according to the in-depth use and the accumulation of feedback, reducing manual intervention and improving efficiency. Through machine learning and intelligent algorithms, the system can dynamically optimize the content display, chart type and data sorting of the report, thereby enhancing the interactivity and adaptability of the report. Ultimately, users can obtain more accurate and efficient reports based on the system's self-optimization and personalized suggestions, thereby providing more valuable support for decision-making. This process not only improves the user experience, but also makes reports more efficient and intelligent in a dynamically changing business environment.

[0167] In this application, based on the generated reports and analysis results, the system can also automatically provide decision-making recommendations and automatically execute some decisions based on preset rules or machine learning models. For example, if inventory data shows that the sales volume of certain drugs continues to be sluggish, the system can automatically recommend reducing the inventory of the drug or adjusting the pricing strategy, and automatically perform the corresponding adjustments (such as updating inventory, initiating purchase requests, etc.) after user confirmation. In this way, by automating some operations, the execution of decisions can be improved and human delays can be reduced.

[0168] Based on generated reports and data analysis results, the system automatically provides decision recommendations and, in some cases, automatically executes them. This process, through pre-set rules and machine learning models, enables the system to quickly make optimization recommendations and automatically execute certain decisions upon user confirmation. This improves response time, reduces human intervention, and ensures that companies can quickly respond to market changes and business needs.

[0169] First, the system conducts in-depth analysis based on the generated report data to identify potential issues. For example, the system analyzes sales data, inventory data, market trends, and other data to automatically detect outliers or potential risks. If the system detects persistently low sales of a particular drug, inventory backlogs, or a marketing campaign failing to achieve expected results, it automatically flags these as potential issues and prepares to generate appropriate decision-making recommendations.

[0170] After identifying a problem, the system automatically generates appropriate decision recommendations based on data analysis results, combined with pre-set decision rules or machine learning models. For example, if sales of certain medications in inventory are sluggish, the system may automatically recommend reducing inventory or adjusting prices based on market demand. Furthermore, the system can also provide users with multiple decision options based on factors such as drug sales trends and market demand fluctuations, helping them make the optimal decision among multiple options.

[0171] Once a user confirms a decision, the system automatically executes portions of the decision based on pre-set rules. For example, if a decision is made to adjust inventory or pricing, the system automatically updates inventory data, adjusts pricing strategies, and even initiates purchase requests or adjusts supply chain orders. This process reduces the tediousness of manual operations and ensures that decisions are implemented quickly. By automating decision execution, the system improves decision execution, reduces human delays, and increases responsiveness. Automated decision execution can help companies adjust operational strategies in real time, especially in rapidly changing market environments.

[0172] After a decision is executed, the system continuously tracks and monitors its effectiveness. For example, it tracks sales after inventory adjustments and market reactions to price adjustments, providing timely feedback on the effectiveness of the decision and further adjusting the decision based on this feedback. If a decision fails to achieve the expected results, the system automatically notifies you and provides suggestions for further improvement. Through this continuous performance monitoring and feedback mechanism, the system can automatically adjust the decision-making process based on actual results, ensuring that each decision is continuously optimized and improving its accuracy and effectiveness.

[0173] Through the intelligent decision-making recommendations and automated execution steps of this invention, enterprises can improve decision-making efficiency and response speed. The system not only identifies potential business issues but also generates accurate decision-making recommendations based on preset rules and machine learning models, and automatically executes decisions after user confirmation. This process reduces human intervention and ensures rapid decision implementation, especially in a dynamically changing market environment, allowing enterprises to adjust strategies more promptly. Through continuous decision feedback and effect monitoring, the system can also continuously optimize the decision-making process, ensuring that each decision is more scientific and effective.

[0174] In this application, during the report generation process, data can also be divided into multiple clusters (groups) through clustering algorithms to identify similar data sets. For example, you can group drugs according to their sales patterns, customer behavior, etc. to find out which drugs are in the best-selling category and which are in a slow-selling state. Clustering algorithms include K-means, DBSCAN, and hierarchical clustering algorithms. These algorithms can classify drugs, customers, or market areas based on the inherent structure in the data, allowing pharmacy management to make targeted inventory adjustments and market strategies.

[0175] By using regression analysis, we can predict the future sales of a particular drug and adjust inventory in advance. Regression algorithms include linear regression, ridge regression, and decision tree regression. By building predictive models, these algorithms enable companies to anticipate sales trends, inventory requirements, and market fluctuations.

[0176] Classification algorithms classify data into different categories and predict the data's category based on known features. In the pharmaceutical sales scenario, classification algorithms can be used to identify which drugs will become bestsellers and which will be slow-selling. For example, support vector machines, random forests, and K-nearest neighbors are suitable classification algorithms. These algorithms use trained models to classify drugs based on factors such as sales potential and market demand, helping companies implement more precise inventory management and marketing strategies. Time series analysis can be used to predict drug sales over the next few months, thereby adjusting procurement plans.

[0177] Association rule learning identifies relationships between variables in data, making it particularly useful for discovering purchasing patterns between medications. For example, association rule algorithms can identify which medications are frequently purchased together, providing a basis for bundled promotions. Association rule algorithms can analyze customer purchasing behavior, uncover potential connections between medications, and guide pharmacies in implementing product combination marketing and cross-selling strategies. The above method can generate the report shown in Table 1 below.

[0178] Table 1 First quarter operating report

[0179]

[0180]

[0181] Figure 4 This is a structural diagram of report customization according to an embodiment of the present invention. Figure 4 As shown, according to the method shown in the above embodiment, Figure 4 The structure of the report is shown. Personalized reports may include a report title and summary, sales data analysis, inventory data analysis, compliance checks, optimization suggestions, charts, and visualizations. Sales data analysis includes total sales, sales volume, and sales data by region; inventory data analysis includes total inventory, inventory turnover, and inventory percentage; compliance checks include data compliance, industry standards checks, and adherence to policy requirements; optimization suggestions include sales strategy optimization, inventory adjustment suggestions, and promotion strategy improvements; charts and visualizations include bar charts, pie charts, and line charts. This application provides a report generation device for medical data. Figure 5 FIG. 1 is a schematic diagram of the structure of a report generating device for medical data according to an embodiment of the present invention. Figure 5 As shown, the report customization device based on medical data includes:

[0182] A data acquisition unit 50 is used to acquire pharmaceutical sales data from multiple pharmaceutical industry data sources;

[0183] The data processing unit 51 is used to perform data cleaning and standardization processing on the pharmaceutical sales data to obtain standardized data after removing redundant data, missing data, and inaccurate data;

[0184] A feature extraction unit 52 is configured to generate structured data by performing multi-dimensional analysis and feature extraction on the standardized data;

[0185] The report generation unit 53 is used to match the structured data with the template fields in the user demand report template, map the structured data to the user demand report template, and generate a personalized report in combination with the compliance check and risk assessment rules embedded in the user demand report template.

[0186] It can be understood that the report generation device for medical data of the present invention can refer to the above report customization method based on medical data.

[0187] See also Figure 6 , an embodiment of the present application further provides an electronic device 600, the electronic device comprising:

[0188] at least one processor; and,

[0189] A memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the report customization method based on medical data in the aforementioned method embodiment.

[0190] An embodiment of the present application also provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the report customization method based on medical data in the aforementioned method embodiment.

[0191] like Figure 6 As shown, the electronic device 60 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. Various programs and data required for the operation of the electronic device 600 are also stored in the RAM 603. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0192] The following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touchpad, a key, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows the electronic device 600 with various devices, it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0193] The above description is merely a specific embodiment of the present disclosure, but the scope of protection of the present disclosure is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this disclosure should be included in the scope of protection of the present disclosure. Therefore, the scope of protection of the present disclosure should be based on the scope of protection of the claims.

Claims

1. A report customization method based on medical data, characterized in that: include: Obtain pharmaceutical sales data from multiple pharmaceutical industry data sources; Cleaning and standardizing the pharmaceutical sales data to obtain standardized data after removing redundant data, missing data, and inaccurate data; Generating structured data by performing multi-dimensional analysis and feature extraction on the standardized data; Based on matching the structured data with the template fields in the user demand report template, the structured data is mapped to the user demand report template, and combined with the compliance check and risk assessment rules embedded in the user demand report template, a personalized report is generated.

2. The report customization method based on medical data according to claim 1, characterized in that: The generating of structured data by performing multi-dimensional analysis and feature extraction on the standardized data includes: performing a correlation analysis between different data fields in the standardized data based on a Pearson correlation coefficient, a Spearman rank correlation coefficient, and / or a point-biserial correlation coefficient to obtain correlation analysis results between the different data fields; By performing principal component analysis on the correlation analysis results between different data fields in the standardized data, characteristic fields are extracted where the correlation degree between different data fields in the standardized data exceeds a predetermined value; The structured data is generated based on a characteristic field in which the degree of correlation between different data fields in the standardized data exceeds a predetermined value.

3. The report customization method based on medical data according to claim 2, characterized in that: The extracting of characteristic fields whose correlation degree between different data fields in the standardized data exceeds a predetermined value by performing principal component analysis on the correlation analysis results between different data fields in the standardized data comprises: If principal component analysis is performed on the correlation analysis results between different data fields in the standardized data, and it is determined that the degree of correlation between the target field and more than half of the different data fields in the standardized data exceeds a predetermined value, the target field is extracted; wherein the target field is one of the data fields in the standardized data.

4. The report customization method based on medical data according to claim 1, characterized in that: The matching of the structured data with the template fields in the user demand report template, mapping the structured data to the user demand report template, and generating a personalized report in combination with the compliance check and risk assessment rules embedded in the user demand report template, includes: If there is field data in the structured data that does not comply with the compliance check and risk assessment rules, then the drugs associated with the fields that do not comply with the compliance check and risk assessment rules are marked as potential risks; The structured data mapping is performed based on the fields marked with potential risks to generate a personalized report with potential risk prompts.

5. The report customization method based on medical data according to claim 2, characterized in that: The correlation analysis between different data fields in the standardized data is performed based on the Pearson correlation coefficient, the Spearman rank correlation coefficient and / or the point-biserial correlation coefficient to obtain the correlation analysis results between different data fields, including: Determine the linear correlation between continuous variables in the standardized data using the Pearson correlation coefficient to obtain a first correlation analysis result between data fields; Determining the monotonic relationship between the non-normally distributed data in the standardized data by using the Spearman rank correlation coefficient to obtain a second correlation analysis result between the data fields; The correlation between continuous variables and binary variables was determined by point-biserial correlation coefficient, and the third correlation analysis results between data fields were obtained; Based on the first correlation analysis result, the second correlation analysis result and the third correlation analysis result, a correlation analysis result between the different data fields is obtained.

6. The report customization method based on medical data according to claim 4, characterized in that: Before generating a personalized report based on the compliance check and risk assessment rules embedded in the user demand report template, the method includes: By using an autoencoder to learn a low-dimensional representation of the data, field data in the structured data that does not comply with compliance inspection and risk assessment rules is identified; wherein the compliance inspection and risk assessment rules include at least one of the following: industry regulations, internal corporate regulations, a reasonable range of market investment, an expected range of seasonal fluctuations, and an expected standard for drug sales.

7. The method for customizing a report based on medical data according to any one of claims 1 to 6, characterized in that: include: When generating personalized reports, a sales trend analysis is performed on the drug sales time series data in the standardized data based on a convolutional neural network and / or a recurrent neural network to extract the periodic variation characteristics of sales peaks and valleys, and the sales trend of the drug within a predetermined future time is predicted based on the periodic variation characteristics of sales peaks and valleys; The generation of a personalized report based on the compliance check and risk assessment rules embedded in the user demand report template includes: Based on the compliance check and risk assessment rules embedded in the user demand report template, the forecast results of the sales trend of the drug within a predetermined time in the future are added to generate a personalized report.

8. The method for customizing a report based on medical data according to any one of claims 1 to 6, characterized in that: The pharmaceutical sales data obtained from multiple pharmaceutical industry data sources includes: Monitor data source status in real time and dynamically adjust the collection priorities of multiple pharmaceutical industry data sources; Based on the collection priorities of the multiple pharmaceutical industry data sources, real-time collection of pharmaceutical sales data is performed.

9. The method for customizing a report based on medical data according to any one of claims 1 to 6, characterized in that: The pharmaceutical sales data is cleaned and standardized to obtain standardized data after removing redundant data, missing data, and inaccurate data, including: Use the rule engine to detect and repair pharmaceutical sales data with missing fields, format errors, and abnormal values; Through the standardization module, the data format of the repaired pharmaceutical sales data is unified and the timestamps are aligned to obtain standardized data with time consistency.

10. A report customization device based on medical data, characterized in that: include: A data acquisition unit, used to acquire pharmaceutical sales data from multiple pharmaceutical industry data sources; A data processing unit, configured to perform data cleaning and standardization on the pharmaceutical sales data to obtain standardized data after removing redundant data, missing data, and inaccurate data; A feature extraction unit, configured to generate structured data by performing multi-dimensional analysis and feature extraction on the standardized data; A report generation unit is used to match the structured data with the template fields in the user demand report template, map the structured data to the user demand report template, and generate a personalized report in combination with the compliance check and risk assessment rules embedded in the user demand report template.

Images