Enterprise intelligent purchase management method and system based on big data

Abstract

The invention discloses an enterprise intelligent purchase management method and system based on big data, and relates to the technical field of enterprise purchase management. The method comprises the following steps: collecting historical purchase data; preprocessing the historical purchase data to obtain preprocessed historical purchase data; key purchase data features are extracted from the preprocessed historical purchase data; the key purchase data features comprise numerical value type features and non-numerical value type features; inputting the numerical value type features and the non-numerical value type features into a clustering analysis model, and performing clustering analysis on purchase data corresponding to the data source to obtain a plurality of preset purchase groups; obtaining current purchase data of the target enterprise; preprocessing the current purchase data to obtain preprocessed current purchase data; and based on the preprocessed current purchase data, determining a target purchase group corresponding to the target enterprise from a plurality of preset purchase groups. By implementing the technical scheme of the invention, the overall efficiency and effect of enterprise purchase management can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of enterprise procurement management, and specifically to an enterprise intelligent procurement management method and system based on big data. Background Art

[0002] In the procurement management of modern enterprises, procurement data usually comes from multiple different data sources, including ERP system, supplier management system, financial system, etc. The procurement information recorded in these data sources varies in form and has complex data types.

[0003] Traditional corporate procurement management methods usually integrate and analyze procurement data manually using simple statistical analysis methods to achieve procurement management. However, this method often has the following shortcomings: first, manual data processing is inefficient and prone to human errors; second, simple statistical analysis methods are difficult to fully explore the deep information in the data and cannot accurately identify the needs of different procurement groups, resulting in poor overall efficiency and effectiveness of corporate procurement management.

[0004] Therefore, how to improve the overall efficiency and effectiveness of enterprise procurement management has become a technical problem that needs to be solved urgently. Summary of the invention

[0005] The present application provides an enterprise intelligent procurement management method and system based on big data, which can improve the overall efficiency and effectiveness of enterprise procurement management.

[0006] In a first aspect, the present application provides an enterprise intelligent procurement management method based on big data, the method comprising: collecting historical procurement data from various data sources related to enterprise procurement; preprocessing each of the historical procurement data to obtain the preprocessed historical procurement data corresponding to each of the data sources; extracting key procurement data features corresponding to each of the preprocessed historical procurement data; the key procurement data features include numerical type features and non-numerical type features; inputting each of the numerical type features and each of the non-numerical type features into a preset clustering analysis model, performing cluster analysis on the procurement data corresponding to each of the data sources, and obtaining multiple preset procurement groups; obtaining the current procurement data of the target enterprise; preprocessing the current procurement data to obtain the preprocessed current procurement data; based on the preprocessed current procurement data, determining the target procurement group corresponding to the target enterprise from the multiple preset procurement groups.

[0007] By adopting the above technical solution, historical procurement data is collected from multiple data sources to ensure the comprehensiveness and diversity of the data, thereby providing a rich information basis for subsequent analysis. By preprocessing the historical procurement data, including processing missing values ​​and error values, the accuracy and completeness of the data are ensured, thereby improving the reliability of data analysis. By extracting key procurement data features, including numerical type features and non-numerical type features, the core information in the procurement behavior is captured, providing effective input data for subsequent cluster analysis. By inputting the extracted features into the cluster analysis model, cluster analysis is performed to identify multiple preset procurement groups with similar procurement behaviors, providing a basis for cluster management for enterprises. By preprocessing the current procurement data, the format and quality of the data are ensured to be consistent with the historical data, thereby ensuring the accuracy and consistency of subsequent analysis. By matching the preprocessed current procurement data with the preset procurement group, the target procurement group of the target enterprise is determined, enabling the enterprise to accurately manage and optimize strategies based on similar procurement behaviors, thereby improving the overall efficiency and effectiveness of the enterprise's procurement management.

[0008] Optionally, the preprocessed historical purchase data includes preprocessed historical purchase data of numerical type and preprocessed historical purchase data of non-numerical type; the preprocessing of each of the historical purchase data to obtain the preprocessed historical purchase data corresponding to each of the data sources specifically includes: pre-identifying each of the historical purchase data to determine the data type and preprocessing type corresponding to each of the historical purchase data; the data type includes numerical type and non-numerical type; the preprocessing type includes missing value processing and error value processing; based on the missing value processing, and / or the error value processing, each of the numerical type of historical purchase data is preliminarily processed to obtain a plurality of preprocessed historical purchase data of numerical type; based on the missing value processing, and / or the error value processing, each of the non-numerical type of historical purchase data is preliminarily processed to obtain a plurality of preprocessed historical purchase data of non-numerical type.

[0009] By adopting the above technical solution, each historical procurement data is pre-identified to determine the data type and preprocessing type, so that appropriate preprocessing strategies can be formulated for different types of data to ensure the pertinence and effectiveness of data processing. By performing missing value processing and error value processing on historical procurement data of numerical type, the data can be repaired and completed, the accuracy and completeness of the data can be improved, and the quality of numerical data can be ensured. By performing missing value processing and error value processing on historical procurement data of non-numerical type, the data can be corrected and completed, the completeness and accuracy of non-numerical data can be improved, and the reliability of data analysis can be ensured.

[0010] Optionally, the numerical type features include a first statistical feature, a first time feature, a first aggregation feature, and a first derivative feature; and the non-numerical type features include a first classification feature and a first text feature.

[0011] Optionally, the step of inputting each of the numerical type features and each of the non-numerical type features into a preset cluster analysis model, performing cluster analysis on the purchase data corresponding to each of the data sources, and obtaining a plurality of preset purchase groups, specifically includes: standardizing each of the first statistical features, each of the first time features, each of the first aggregation features, and each of the first derivative features to obtain a plurality of second statistical features, a plurality of second time features, a plurality of second aggregation features, and a plurality of second derivative features; using one-hot encoding to convert each of the first classification features into a binary vector to obtain a plurality of second classification features; and using the corresponding numerical form to perform cluster analysis on each of the first text features. Indicates that a plurality of second text features are obtained; a plurality of the second statistical features, a plurality of the second time features, a plurality of the second aggregation features, a plurality of the second derivative features, a plurality of the second classification features and a plurality of the second text features are integrated into a unified data frame, wherein the data frame includes a plurality of data points, each row of the data frame represents a purchase record, and each column of the data frame represents a feature; determines a clustering algorithm and clustering parameters; inputs the data frame into the clustering algorithm, and performs training based on the clustering parameters to obtain a cluster label corresponding to each of the data points; based on each of the cluster labels, obtains a plurality of the preset purchase groups.

[0012] By adopting the above technical solution, the first statistical feature, the first time feature, the first aggregation feature and the first derivative feature are standardized, thereby eliminating the dimensional differences between the features, improving the comparability of the features in cluster analysis, and ensuring the accuracy and consistency of the analysis results. By using unique hot encoding to convert the first classification feature into a binary vector, the classification feature can be effectively used in numerical calculations, and the processing ability of cluster analysis for classification information is improved. By expressing the first text feature in a corresponding numerical form, the text feature can be effectively used in cluster analysis, and the processing and analysis ability of text information is improved. By integrating all processed features into a unified data framework, it is ensured that each data point contains complete feature information, thereby providing a high-quality data input for subsequent cluster analysis. By selecting an appropriate clustering algorithm and optimizing clustering parameters, the effect and accuracy of cluster analysis can be ensured, and different purchasing groups can be better identified. By inputting the data framework into the clustering algorithm for training, the data points are assigned to different clusters according to feature similarity, and the cluster label of each data point is obtained, so as to achieve effective classification of procurement records. By analyzing and summarizing each cluster label, multiple preset purchasing groups with similar purchasing behaviors can be identified, thereby providing enterprises with accurate purchasing management strategies and improving the efficiency and effectiveness of purchasing management.

[0013] Optionally, determining the target purchasing group corresponding to the target enterprise from the plurality of preset purchasing groups based on the preprocessed current purchasing data specifically includes: extracting a third statistical feature, a third time feature, a third aggregation feature, a third derivative feature, a third classification feature and a third text feature from the preprocessed current purchasing data; calculating the statistical feature similarity between the third statistical feature and each of the second statistical features, and calculating the time feature similarity between the third time feature and each of the second time features, and calculating the aggregation feature similarity between the third aggregation feature and each of the second aggregation features, and calculating the derivative feature similarity between the third derivative feature and each of the second derivative features, and calculating the classification feature similarity between the third classification feature and each of the second classification features, and calculating the text feature similarity between the third text feature and each of the second text features; performing weighted summation on the statistical feature similarity, the time feature similarity, the aggregation feature similarity, the derivative feature similarity, the classification feature similarity and the text feature similarity corresponding to each of the preset purchasing groups to obtain a total similarity score corresponding to each of the preset purchasing groups; and selecting the preset purchasing group with the highest total similarity score as the target purchasing group.

[0014] By adopting the above technical solution, the third statistical feature, the third time feature, the third aggregation feature, the third derivative feature, the third classification feature and the third text feature are extracted from the pre-processed current procurement data, thereby ensuring that the features of the current procurement data are consistent with the features of the historical data, and providing a basis for subsequent similarity calculation and group matching. By calculating the similarity between the third statistical feature, the third time feature, the third aggregation feature, the third derivative feature, the third classification feature and the third text feature and each second feature, the feature similarity between the current procurement data and each preset procurement group is quantified, providing a data basis for further similarity synthesis. By weighted summing the statistical feature similarity, the time feature similarity, the aggregation feature similarity, the derivative feature similarity, the classification feature similarity and the text feature similarity, the importance of different features is comprehensively considered, and the total similarity score of each preset procurement group is obtained, thereby achieving comprehensive quantification of feature similarity. By selecting the preset procurement group with the highest total similarity score as the target procurement group, it is ensured that the selected group has the highest similarity with the current procurement data, and accurate procurement behavior matching is achieved.

[0015] Optionally, the method also includes: calculating the statistical feature similarity, the time feature similarity and the aggregation feature similarity based on the Euclidean distance or cosine similarity calculation method; calculating the classification feature similarity based on the Jaccard similarity or cosine similarity calculation method; calculating the text feature similarity based on the TF-IDF cosine similarity calculation method.

[0016] Optionally, after determining the target procurement group corresponding to the target enterprise from multiple preset procurement groups based on the preprocessed current procurement data, the method further includes: determining a preset management plan corresponding to the target procurement group; wherein the preset management plan includes a procurement strategy management plan, a supplier management plan, a procurement plan and execution plan, an inventory management plan, a cost control and optimization plan, and a risk management plan; and sending the preset management plan to the management terminal of the target enterprise so that the target enterprise performs procurement management according to the preset management plan.

[0017] In a second aspect of the present application, an enterprise intelligent procurement management system based on big data is provided, the system comprising: an acquisition module, a preprocessing module, a feature extraction module, a clustering module and a procurement group determination module; the acquisition module is used to collect historical procurement data from various data sources related to enterprise procurement; the preprocessing module is used to preprocess each of the historical procurement data to obtain the preprocessed historical procurement data corresponding to each of the data sources; the feature extraction module is used to extract key procurement data features corresponding to each of the data sources from each of the preprocessed historical procurement data; the key procurement data features include numerical type features and non-numerical type features; the clustering module is used to input each of the numerical type features and each of the non-numerical type features into a preset clustering analysis model, perform clustering analysis on the procurement data corresponding to each of the data sources, and obtain multiple preset procurement groups; the acquisition module is also used to acquire the current procurement data of the target enterprise; the preprocessing module is also used to preprocess the current procurement data to obtain the preprocessed current procurement data; the procurement group determination module is used to determine the target procurement group corresponding to the target enterprise from the multiple preset procurement groups based on the preprocessed current procurement data.

[0018] In the third aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes any method as described in the first aspect of the present application.

[0019] In a fourth aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program that can be loaded by a processor and execute the method as described in any one of the first aspects of the present application.

[0020] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: By collecting historical procurement data from multiple data sources, the comprehensiveness and diversity of the data are ensured, thereby providing a rich information basis for subsequent analysis. By preprocessing the historical procurement data, including processing missing values ​​and error values, the accuracy and completeness of the data are ensured, thereby improving the reliability of data analysis. By extracting key procurement data features, including numerical type features and non-numerical type features, the core information in the procurement behavior is captured, providing effective input data for subsequent cluster analysis. By inputting the extracted features into the cluster analysis model, cluster analysis is performed to identify multiple preset procurement groups with similar procurement behaviors, providing a basis for cluster management for enterprises. By preprocessing the current procurement data, the format and quality of the data are ensured to be consistent with the historical data, thereby ensuring the accuracy and consistency of subsequent analysis. By matching the preprocessed current procurement data with the preset procurement group, the target procurement group of the target enterprise is determined, enabling the enterprise to accurately manage and optimize strategies based on similar procurement behaviors, thereby improving the overall efficiency and effectiveness of the enterprise's procurement management. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is one of the flow charts of an enterprise intelligent procurement management method based on big data provided in an embodiment of the present application; Figure 2 This is the second flow chart of an enterprise intelligent procurement management method based on big data provided in an embodiment of the present application; Figure 3 This is a flowchart of a method for intelligent procurement management of an enterprise based on big data provided by an embodiment of the present application; Figure 4 This is a structural diagram of an enterprise intelligent procurement management system based on big data provided by an embodiment of the present application; Figure 5 It is a structural schematic diagram of an electronic device disclosed in an embodiment of the present application.

[0022] Explanation of the accompanying drawings: 1. Acquisition module; 2. Preprocessing module; 3. Feature extraction module; 4. Clustering module; 5. Purchasing group determination module; 6. Management module; 500. Electronic device; 501. Processor; 502. Communication bus; 503. User interface; 504. Network interface; 505. Memory. DETAILED DESCRIPTION

[0023] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0024] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0025] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" can explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0026] This application provides an enterprise intelligent procurement management method based on big data, referring to Figure 1 , which shows one of the flow diagrams of a method for enterprise intelligent procurement management based on big data provided in an embodiment of the present application. The method includes steps S1-S7, which are as follows: Step S1: Collect historical procurement data from various data sources related to enterprise procurement.

[0027] Specifically, in modern enterprise procurement management, the integrity and comprehensiveness of data are the basis for optimizing procurement decisions and management. In order to fully understand the procurement behavior of an enterprise, historical procurement data must be collected from multiple data sources related to enterprise procurement. These data sources include but are not limited to enterprise resource planning (ERP) systems, supplier management systems, financial systems, and other related databases.

[0028] During the specific implementation process, we first need to determine the type and access method of each data source. For example, the ERP system usually records detailed purchase orders, inventory information, purchase invoices and other data, the supplier management system contains information such as supplier evaluation, contract management and delivery records, and the financial system records financial data such as procurement costs, payment status and budget execution.

[0029] In the process of collecting data, a variety of technical means can be used, such as database connection, API call and file import, to ensure accurate data extraction and efficient integration. In order to ensure the quality and consistency of data, it is necessary to analyze the data format and structure of different data sources and design corresponding data conversion and cleaning processes. For example, for purchase order data in the ERP system, the required fields can be obtained through SQL query and converted into a standardized format; for contract data in the supplier management system, it can be extracted through API calls and parsed into structured data; for payment records in the financial system, financial report files can be imported and data parsed and cleaned.

[0030] Through these steps, historical procurement data collected from various data sources will be integrated into a unified database to form a comprehensive procurement data warehouse. This data warehouse not only contains numerical data such as procurement amount, quantity, time, etc., but also contains non-numerical data such as supplier information and contract text. This multi-source data integration method can provide a panoramic view of corporate procurement behavior and provide a rich and reliable data foundation for subsequent data preprocessing, feature extraction and cluster analysis.

[0031] Step S2: pre-process each historical purchase data to obtain the pre-processed historical purchase data corresponding to each data source.

[0032] Specifically, in enterprise procurement management, historical procurement data collected from various data sources often have problems such as inconsistent formats, missing data, and incorrect values. In order to ensure the accuracy and consistency of the data, these historical procurement data must be preprocessed. Through preprocessing, data quality can be improved and the reliability of subsequent analysis and decision-making can be ensured.

[0033] Since the historical purchase data can be divided into numerical historical purchase data and non-numerical historical purchase data, the preprocessed historical purchase data includes the preprocessed numerical historical purchase data and the preprocessed non-numerical historical purchase data.

[0034] In a possible implementation, step S2 specifically includes the following steps: Pre-identify each historical procurement data to determine the data type and preprocessing type corresponding to each historical procurement data; the data type includes numeric type and non-numeric type; the preprocessing type includes missing value processing and error value processing.

[0035] Specifically, the pre-identification process includes the following key links. First, scan and classify each historical procurement data to identify its data type. Data types are mainly divided into numerical types and non-numerical types. Numerical types include procurement amount, quantity, time, etc., while non-numerical types include supplier name, procurement category, contract text, etc. Through automated tools or scripts, the type of each data field can be quickly identified and labeled.

[0036] Next, based on the preprocessing requirements, determine the preprocessing type for each data field. Preprocessing types include missing value processing and error value processing. Missing value processing is to fill in the gaps in the data to avoid the impact of incomplete data on the analysis results. Error value processing is to correct anomalies and errors in the data to ensure the accuracy of the data. For example, if there are missing values ​​in the purchase amount field, missing value filling is required; if the purchase quantity is found to be negative, error value correction is required.

[0037] Based on missing value processing and / or error value processing, historical purchase data of each numerical type are preliminarily processed to obtain multiple preprocessed historical purchase data of numerical types.

[0038] Specifically, we first need to process missing values ​​for historical purchase data of numerical type. Missing values ​​can lead to inaccurate data analysis, so we need to choose an appropriate method to fill them. Common missing value processing methods include mean filling, median filling, and interpolation. For numerical data such as purchase amount and purchase quantity, we can use mean filling or median filling to ensure that the filled data does not deviate from the overall data distribution. For example, if the purchase amount in some purchase records is missing, we can calculate the average purchase amount of the data set and use the average value to fill the missing values.

[0039] Next, the historical procurement data of numerical type is processed for error values. Error values ​​may be outliers caused by data entry errors, system failures, or other reasons. If these outliers are not handled, they will have a significant impact on the results of data analysis. Common error value processing methods include outlier detection and logical checks. Outlier detection can be identified and processed using statistical methods (such as z-score, IQR). For example, by calculating the standard deviation of the data, values ​​that exceed a certain standard deviation range can be considered outliers and corrected to values ​​within a reasonable range. Logical checks are performed according to business rules. For example, the purchase quantity should not be negative. If a negative value is found, it is corrected to a reasonable positive value.

[0040] In the specific implementation process, Python's Pandas library can be used to implement data processing. For missing value processing, the fillna method can be used to fill the mean or median; for error value processing, the apply method combined with a custom function can be used to detect and correct outliers.

[0041] Based on missing value processing and / or error value processing, each non-numeric type of historical procurement data is preliminarily processed to obtain a plurality of pre-processed non-numeric type of historical procurement data.

[0042] Specifically, we first need to process missing values ​​for non-numeric historical procurement data. Missing values ​​will lead to incomplete information, which will affect the effect of data analysis. For categorical data (such as supplier name, procurement category, etc.), we can use the most frequent value filling method, that is, fill the missing values ​​with the most common value in the field. This can maintain the distribution characteristics of the data and reduce the impact of filling on data analysis. For example, if the supplier name in some procurement records is missing, we can calculate the supplier name with the highest frequency in the data set and use it to fill the missing values.

[0043] For text data (such as contract text, purchase notes, etc.), missing value processing can choose to delete records containing missing values ​​or fill them with default values ​​according to the specific situation. If the missing text data has little impact on the overall analysis, you can choose to delete these records; if there are many missing values, you can fill them with default values ​​such as "no notes" to maintain the integrity of the data set.

[0044] Next, handle the error values ​​of non-numeric historical purchase data. Error values ​​may be unexpected values ​​due to data entry errors or other reasons. If these error values ​​are not handled, they will have a significant impact on the data analysis results. For categorical data, you can process it through data validation rules. For example, check whether the value in the purchase category field is in the predefined category list. If a value is found that is not in the list, it can be replaced with the closest correct value or marked as "unknown category".

[0045] For text data error value processing, natural language processing (NLP) technology can be used to correct it. For example, spelling errors in contract text can be detected and corrected through spelling check tools, or format errors in text can be identified and corrected through text matching algorithms.

[0046] In the specific implementation process, Python's Pandas library and NLP tools can be used for data processing. For missing value processing, the fillna method can be used to fill the most frequent value; for error value processing, the apply method combined with a custom function can be used to verify and correct the data.

[0047] Step S3: extracting key procurement data features corresponding to each data source from each preprocessed historical procurement data; the key procurement data features include numerical type features and non-numerical type features.

[0048] Specifically, in the analysis of corporate procurement data, the pre-processed historical procurement data needs to further extract key procurement data features in order to provide effective input data for subsequent cluster analysis. The purpose of extracting these features is to grasp the core information in procurement behavior, thereby optimizing procurement decisions and improving the efficiency and effectiveness of procurement management.

[0049] Among them, the numerical type features include a first statistical feature, a first time feature, a first aggregation feature, and a first derivative feature; the non-numerical type features include a first classification feature and a first text feature.

[0050] First, for historical procurement data of numerical type, it is crucial to extract statistical features, time features, aggregate features, and derived features. Statistical features include the mean, standard deviation, minimum, and maximum values ​​of data such as procurement amount and procurement quantity. These features can describe the central tendency and dispersion of data. For example, by calculating the mean and standard deviation of the procurement amount, we can understand the average procurement amount and its fluctuation. Time features reflect the distribution and changes of procurement behavior in the time dimension, such as procurement time point, time interval, and periodicity. These features help identify seasonal and periodic patterns of procurement. By analyzing the time interval between each procurement, we can find the frequency and regularity of procurement. Aggregate features are an overall description of the data, such as total procurement amount, total procurement quantity, etc., which can reflect the total amount of procurement activities and thus measure the scale and importance of procurement. Derived features are advanced features calculated from basic features, such as unit price and growth rate. These features can provide deeper analysis information. Changes in unit price can reflect market price fluctuations.

[0051] Next, for non-numerical historical procurement data, it is equally important to extract classification features and text features. Classification features include supplier names, procurement categories, etc. These features can help analyze the distribution and differences of data in different categories. For example, by analyzing the supplier names, you can understand the differences in procurement behavior of different suppliers. Text features involve contract texts, procurement notes, etc. These text data can be converted into numerical features through natural language processing technology. Using the TF-IDF (term frequency-inverse document frequency) method, representative keywords and phrases can be extracted for further analysis.

[0052] Step S4: Input each numerical type feature and each non-numerical type feature into a preset cluster analysis model, perform cluster analysis on the purchase data corresponding to each data source, and obtain multiple preset purchase groups.

[0053] Specifically, the extracted features need to be further input into the cluster analysis model to identify different purchasing groups. The purpose of cluster analysis is to group data points with similar purchasing behaviors, thereby identifying different purchasing groups, providing enterprises with targeted management strategies, and improving the efficiency and effectiveness of purchasing management.

[0054] In one possible implementation, refer to Figure 2 , which shows the second flow diagram of an enterprise intelligent procurement management method based on big data provided by an embodiment of the present application. Step S4 specifically includes steps S41-S47: Step S41: performing standardization processing on each first statistical feature, each first time feature, each first aggregate feature, and each first derivative feature to obtain a plurality of second statistical features, a plurality of second time features, a plurality of second aggregate features, and a plurality of second derivative features.

[0055] Specifically, it is necessary to first standardize each numerical type feature, including the first statistical feature, the first time feature, the first aggregation feature, and the first derived feature. The first statistical feature, such as the mean and standard deviation of the purchase amount, the first time feature, such as the interval and cycle of the purchase time, the first aggregation feature, such as the total purchase amount and total purchase quantity, and the first derived feature, such as the unit price and growth rate, may have different dimensions and ranges in the original data. Directly using them for cluster analysis will cause data with larger feature values ​​to have an unbalanced impact on the clustering results.

[0056] For standardization, the z-score standardization method can be used. The specific steps are as follows: for each feature, first calculate its mean and standard deviation, then subtract the mean from each data point and divide it by the standard deviation.

[0057] In actual operation, standardization can be easily implemented using the data processing library in Python. For example, the StandardScaler class in the Scikit-learn library can be used to standardize data. First, import the standardization library and instantiate the StandardScaler object. Then, use the fit_transform method to standardize the first statistical feature, the first time feature, the first aggregation feature, and the first derivative feature to obtain multiple second statistical features, multiple second time features, multiple second aggregation features, and multiple second derivative features.

[0058] Step S42: Use one-hot encoding to convert each first classification feature into a binary vector to obtain multiple second classification features.

[0059] In enterprise procurement data analysis, categorical features (such as supplier name, purchase category, etc.) cannot be directly used for numerical calculations and cluster analysis because these features are non-numerical. In order to effectively use these categorical features in cluster analysis models, they must be converted into numerical form. One-Hot Encoding is a commonly used method that converts categorical features into binary vectors so that each category is represented by a separate binary feature, thereby retaining the categorical information and making it suitable for numerical calculations.

[0060] Specifically, one-hot encoding creates a new binary feature for each category. When a data point belongs to a category, the value of the binary feature is 1, and the rest is 0. For example, suppose the supplier name has three different values: "Supplier A", "Supplier B", and "Supplier C". One-hot encoding converts it into three binary features: Supplier_A, Supplier_B, and Supplier_C. If the supplier of a purchase record is "Supplier A", then in the encoded feature, Supplier_A is 1, and Supplier_B and Supplier_C are 0.

[0061] In practice, this conversion process can be achieved by using the one-hot encoding tool in Python's Pandas library and Scikit-learn library. First, import the necessary libraries and create a one-hot encoder object. Then, use the one-hot encoder to convert the first categorical feature to generate multiple second categorical features in the form of binary vectors. Next, add these binary vector features to the original dataset.

[0062] Through one-hot encoding, each first classification feature is successfully converted into multiple second classification features in the form of binary vectors. These binary features not only retain the information of the original classification features, but also enable the classification data to be effectively used in cluster analysis. The converted data structure is clearer and more standardized, ensuring the accuracy and consistency of subsequent analysis.

[0063] For example, suppose a dataset contains a categorical feature called supplier name, which can be converted into multiple binary features through one-hot encoding. During the conversion process, the one-hot encoder creates a binary feature for each supplier and encodes these features according to the actual value of each data point. In the encoded dataset, each data point is represented as a set of binary vectors, so that the categorical information can be directly used for numerical calculations and cluster analysis.

[0064] Step S43: Represent each first text feature in a corresponding numerical form to obtain a plurality of second text features.

[0065] In enterprise procurement data analysis, text features (such as contract text, procurement notes, etc.) contain a lot of valuable information, but due to their unstructured nature, they cannot be directly used for numerical calculations and cluster analysis. In order to effectively use text features in cluster analysis models, they must be converted into numerical form. TF-IDF (Term Frequency-Inverse Document Frequency) is a commonly used method that converts text data into numerical vectors by measuring the importance of words in documents, making them suitable for numerical calculations and cluster analysis.

[0066] Specifically, the TF-IDF method reflects the importance of each word in a document by calculating the product of term frequency and inverse document frequency. Term frequency (TF) indicates the frequency of a word in a document, while inverse document frequency (IDF) measures the rarity of a word in all documents. The higher the TF-IDF value, the more important the word is in the current document. In this way, text features can be converted into a sparse matrix, where each document is represented as a numerical vector, and each dimension of the vector corresponds to the TF-IDF value of a word.

[0067] In practice, this conversion process can be implemented using the TfidfVectorizer class in Python's Scikit-learn library. First, import the necessary libraries and create a TF-IDF vectorizer object. Then, use the TF-IDF vectorizer to fit and transform the first text feature to generate the corresponding TF-IDF matrix. Next, convert the TF-IDF matrix into a data frame and merge it with the original dataset to obtain multiple second text features.

[0068] Through TF-IDF vectorization, the original unstructured text data is successfully converted into a structured numerical form. These numerical features not only retain the important information in the text data, but also enable the text features to be effectively used in clustering analysis. The converted data structure is clearer and more standardized, ensuring the accuracy and consistency of subsequent analysis.

[0069] For example, suppose a dataset contains a text feature called contract text, which can be converted into multiple numerical features through TF-IDF vectorization. During the conversion process, the TF-IDF vectorizer analyzes all documents, calculates the TF-IDF value of each word, and represents each document as a TF-IDF vector. Ultimately, all text features are represented as a sparse matrix, where the numerical representation of each document facilitates subsequent clustering analysis.

[0070] Step S44: Integrate multiple second statistical features, multiple second time features, multiple second aggregation features, multiple second derivative features, multiple second classification features, and multiple second text features into a unified data frame, wherein the data frame includes multiple data points, each row of the data frame represents a purchase record, and each column of the data frame represents a feature.

[0071] Specifically, you first need to convert each feature data into a data frame for merging. For numeric features, you can add them directly to the data frame. For categorical features after one-hot encoding, each category will generate a new binary feature column, which also needs to be added to the data frame. For text features after TF-IDF vectorization, the generated sparse matrix needs to be converted to a data frame and merged with other feature data.

[0072] During the integration process, it is necessary to ensure that each row in the data frame represents a purchase record and each column represents a feature. In this way, it is possible to ensure that each data point contains all the necessary information to form a complete data frame. For example, all processed feature data are merged according to the same index to ensure that the feature data of each purchase record is in the same row.

[0073] In this way, multiple processed feature data are integrated into a unified data framework to form a structured data set. This data framework not only contains all the necessary feature information, but also has a unified structure and form, which is convenient for subsequent clustering analysis and model training.

[0074] Step S45: Determine the clustering algorithm and clustering parameters.

[0075] Specifically, first, you need to select a suitable clustering algorithm based on the characteristics of the data and the analysis requirements. Common clustering algorithms include K-means, Hierarchical Clustering, and DBSCAN. The K-means algorithm is suitable for situations where data points are relatively concentrated and the cluster shape is spherical. It has the advantages of high computational efficiency and easy implementation. The hierarchical clustering algorithm does not require the number of clusters to be specified in advance. It is suitable for situations where data points are unevenly distributed or the cluster shape is complex. The DBSCAN algorithm can recognize clusters of any shape and can automatically identify noise points. It is suitable for situations where noisy data exists.

[0076] After determining the clustering algorithm, you need to set the clustering parameters. Taking the K-means algorithm as an example, the most important parameter is the number of clusters k. The selection of a suitable k value can be evaluated by a variety of methods, including the Elbow Method and the Silhouette Score. The Elbow Method plots a curve of the clustering cost function as the number of clusters changes, and selects the k value at the inflection point of the curve. The Silhouette Score evaluates the clustering effect by calculating the distance between each data point and its cluster and the nearest cluster. The higher the average silhouette coefficient corresponding to the k value, the better the clustering effect.

[0077] In the specific implementation process, Python's Scikit-learn library can be used to implement clustering algorithm selection and parameter optimization. First, import the necessary libraries and create a clustering algorithm object. Then, determine the best parameters through evaluation methods. For example, use the elbow method to draw a clustering cost function curve, or use the silhouette coefficient to evaluate the clustering effect of different k values. Finally, select the optimal clustering algorithm and parameters and prepare for clustering analysis.

[0078] Step S46: Input the data frame into the clustering algorithm and perform training based on the clustering parameters to obtain a cluster label corresponding to each data point.

[0079] Specifically, this data frame is input into the preset clustering algorithm. Assume that the K-means clustering algorithm is selected, and the optimal number of clusters k is determined through preliminary analysis. The K-means algorithm needs to be used to train the data so that it can assign data points to different clusters based on various features. Specifically, the K-means algorithm minimizes the distance from each data point in each cluster to the cluster center through an iterative optimization process, thereby achieving the purpose of clustering data points.

[0080] In practice, Python's Scikit-learn library can be used to implement this process. First, create a K-means clustering object and set the number of clusters k. Then, input the data in the data frame into the clustering algorithm and call K-means' fit method to train the model. The algorithm will iteratively calculate based on the data features, continuously adjust the position of the cluster center, and finally assign each data point to the nearest cluster center. After training, the model will generate a cluster label for each data point, indicating the cluster to which the data point belongs.

[0081] Step S47: Based on each cluster label, multiple preset purchasing groups are obtained.

[0082] Specifically, first, these cluster labels need to be analyzed and summarized to define and describe multiple preset purchasing groups. By counting the data features within each cluster, the mean and standard deviation of each feature can be calculated to understand the central characteristics and degree of variation of each cluster. For example, by calculating the mean and standard deviation of the purchase amount of each cluster, the average purchase scale and fluctuation of the cluster can be understood. At the same time, by analyzing the time characteristics of each cluster, the distribution and regularity of different clusters in the time dimension can be identified to understand whether there are seasonal purchasing patterns or other time-related purchasing behaviors.

[0083] During the analysis, it is also necessary to summarize the distribution of classification features and text features. For each cluster, the distribution of classification features (such as supplier categories, procurement categories, etc.) and the main content and keywords of text features (such as contract text, procurement notes) are counted. This information helps to understand the specific classification features and text information of each cluster and further refine the feature description of each procurement group.

[0084] Through in-depth analysis of cluster labels, these descriptions can be transformed into the definition of preset purchasing groups. Each preset purchasing group contains similar purchasing records with similar characteristics and behavior patterns. For example, by analyzing the characteristic mean and standard deviation of a cluster, it is determined that the purchasing amount of the cluster is large and the fluctuation is small, indicating that the group may be a high-value and stable purchasing group. Another cluster may show a high purchase frequency and obvious seasonal changes, indicating that the group may be a seasonal purchasing group.

[0085] Step S5: Obtain the current purchasing data of the target enterprise.

[0086] Specifically, first, it is necessary to determine the source of current procurement data. The target company's procurement data is usually stored in multiple databases such as the enterprise resource planning (ERP) system, supplier management system, and financial system. In order to fully obtain current procurement data, it is necessary to extract the latest procurement records from these data sources, including purchase orders, supplier information, purchase amount, purchase quantity, delivery time and other detailed information.

[0087] In the specific operation process, a variety of technical means can be used to obtain data. Current procurement data can be extracted from various data sources through database connection, API call or file import. For example, the latest purchase order data can be extracted from the ERP system through SQL query, real-time supplier information can be obtained from the supplier management system using API, or the latest purchase payment records in the financial system can be imported.

[0088] Step S6: pre-process the current purchase data to obtain pre-processed current purchase data.

[0089] Specifically, the specific implementation method of step S6 is consistent with the specific implementation method of step S2, so it will not be described in detail here.

[0090] Step S7: Based on the pre-processed current procurement data, determine the target procurement group corresponding to the target enterprise from multiple preset procurement groups.

[0091] Specifically, by analyzing the pre-processed current procurement data, the target procurement group corresponding to the target enterprise can be determined. The purpose of this step is to match the procurement behavior of the target enterprise with multiple preset procurement groups and identify the group that is most similar to it, so as to provide the enterprise with accurate procurement management strategies and improve procurement efficiency and effectiveness.

[0092] In one possible implementation, refer to Figure 3 , which shows the third flow chart of a method for enterprise intelligent procurement management based on big data provided in an embodiment of the present application. Step S7 specifically includes steps S71-S74: Step S71: extracting a third statistical feature, a third time feature, a third aggregation feature, a third derivative feature, a third classification feature and a third text feature from the preprocessed current procurement data.

[0093] Specifically, the specific implementation method of step S71 is consistent with the specific implementation method of step S3, so it will not be described in detail here.

[0094] Step S72: Calculate the statistical feature similarity between the third statistical feature and each second statistical feature, calculate the time feature similarity between the third time feature and each second time feature, calculate the aggregation feature similarity between the third aggregation feature and each second aggregation feature, calculate the derivative feature similarity between the third derivative feature and each second derivative feature, calculate the classification feature similarity between the third classification feature and each second classification feature, and calculate the text feature similarity between the third text feature and each second text feature.

[0095] In a possible implementation, step S72 specifically includes the following steps: Based on the Euclidean distance or cosine similarity calculation method, the statistical feature similarity, temporal feature similarity and aggregate feature similarity are calculated.

[0096] Calculate the similarity of classification features based on the Jaccard similarity or cosine similarity calculation method.

[0097] Calculate text feature similarity based on the TF-IDF cosine similarity calculation method.

[0098] Specifically, in order to accurately match the current purchasing data of the target enterprise with multiple preset purchasing groups, the similarity between various features must be calculated. Through similarity calculation, the similarity between the current purchasing data and each preset purchasing group can be quantified, thereby determining the most similar purchasing group.

[0099] First, calculate the statistical feature similarity between the third statistical feature and each second statistical feature. Statistical features reflect the basic statistical information of procurement data, such as mean, standard deviation, etc. Commonly used similarity calculation methods include Euclidean distance and cosine similarity. Euclidean distance quantifies the similarity by calculating the straight-line distance between two sets of feature values. The smaller the distance, the higher the similarity. Cosine similarity quantifies the similarity by calculating the cosine value of the angle between two sets of feature values. The larger the cosine value, the higher the similarity. Choosing an appropriate method to calculate the similarity between the third statistical feature and each second statistical feature can accurately measure the similarity between the current procurement data and the preset procurement group in terms of statistical features.

[0100] Next, the time feature similarity between the third time feature and each second time feature is calculated. The time feature reflects the distribution and changes of the purchasing behavior in the time dimension, such as the purchasing frequency, time interval, etc. The similarity between the time features can be calculated using the same Euclidean distance or cosine similarity as the statistical feature. By measuring the similarity of the time features, the degree of match between the current purchasing behavior and the preset purchasing group in the time pattern can be identified.

[0101] Then, the similarity of the third aggregate feature with each second aggregate feature is calculated. The aggregate feature describes the total amount of purchase data, such as the total purchase amount, the total purchase quantity, etc. The similarity between the aggregate features can be quantified by using the Euclidean distance or cosine similarity method to determine the degree of similarity between the current purchase data and the preset purchase group in terms of overall scale.

[0102] At the same time, the derivative feature similarity between the third derivative feature and each second derivative feature is calculated. Derived features are advanced features calculated from basic features, such as unit price, growth rate, etc. By using Euclidean distance or cosine similarity, the similarity between derived features can be accurately measured to identify the degree of match between the current purchasing behavior and the preset purchasing group in terms of derived features.

[0103] For classification features, the classification feature similarity between the third classification feature and each second classification feature is calculated. Classification features include supplier categories, purchase categories, etc. Commonly used similarity calculation methods include Jaccard similarity and cosine similarity. Jaccard similarity quantifies similarity by calculating the ratio of the intersection and union of two sets, and is suitable for measuring the similarity of classification features. By calculating the classification feature similarity, the degree of match between the current procurement data and the preset procurement group in category distribution can be identified.

[0104] Finally, the text feature similarity between the third text feature and each second text feature is calculated. Text features such as contract text and procurement notes can be processed using the TF-IDF cosine similarity method. TF-IDF cosine similarity quantifies the similarity by calculating the cosine value between text vectors. It can accurately measure the similarity between text features and identify the degree of match between the current procurement data and the preset procurement group in text information.

[0105] It should be noted that before calculating the text feature similarity between the third text feature and each second text feature, text features such as contract text, procurement notes, etc. may have been converted into numerical form after TF-IDF processing, so the cosine similarity can also be used directly to calculate the text feature similarity.

[0106] Step S73: weighted summing up the statistical feature similarity, time feature similarity, aggregate feature similarity, derivative feature similarity, classification feature similarity and text feature similarity corresponding to each preset purchasing group to obtain a total similarity score corresponding to each preset purchasing group.

[0107] Specifically, each preset purchasing group is defined based on the second statistical feature, the second time feature, the second aggregation feature, the second derivative feature, the second classification feature, and the second text feature. These features describe the characteristics and behavior patterns of the purchasing group in terms of quantity, time, total volume, derivative characteristics, classification, and text content. For example, a preset purchasing group may have specific patterns in terms of purchase amount (second statistical feature), purchase frequency (second time feature), and supplier category (second classification feature), which are reflected in the definition of the group.

[0108] In order to comprehensively measure the similarity between the current procurement data and these preset procurement groups, it is necessary to perform a weighted sum of the similarities on different feature dimensions. Specifically, a weight needs to be assigned to each feature, and these weights can be set based on the importance of the feature in the procurement behavior. After setting the weights, the weighted sum of the similarities of various features for each preset procurement group is performed. The specific formula is as follows: Total similarity score = w1×statistical feature similarity + w2×time feature similarity + w3×aggregate feature similarity + w4×derivative feature similarity + w5×classification feature similarity + w6×text feature similarity, where w1, w2, w3, w4, w5, and w6 represent the weights of each feature, respectively.

[0109] By weighted summation, the total similarity score of each preset purchasing group can be obtained. These scores comprehensively reflect the overall similarity between the current purchasing data and each preset purchasing group in each characteristic dimension. The higher the score, the higher the similarity between the current purchasing data and the preset purchasing group.

[0110] Step S74: Select the preset purchasing group with the highest total similarity score as the target purchasing group.

[0111] Specifically, the total similarity scores of all preset purchasing groups are traversed to find the group with the highest score. The reason for selecting this group is that the characteristics and behavior patterns of this group are closest to the current purchasing data of the target enterprise, so it can provide the most valuable reference and management suggestions.

[0112] For example, through calculation, we know that a certain preset purchasing group A has the highest total similarity score, which means that the current purchasing behavior of the target enterprise is most similar to the characteristics of group A in multiple dimensions. This similarity is not only reflected in statistical features, such as the distribution of purchase amount and quantity, but also includes the purchase frequency and time interval in time features, the total purchase amount and quantity in aggregate features, the unit price and growth rate in derived features, and the matching of classification features and text features.

[0113] In a possible implementation, after step S7, the method further includes the following steps: Determine the preset management plan corresponding to the target purchasing group; among them, the preset management plan includes procurement strategy management plan, supplier management plan, procurement plan and execution plan, inventory management plan, cost control and optimization plan and risk management plan.

[0114] Specifically, in corporate procurement management, after identifying and determining the target procurement groups, companies need to develop detailed preset management plans based on the characteristics and behavior patterns of these groups to optimize the procurement process and improve management efficiency. By analyzing the characteristics of the target procurement groups, such as procurement amount, procurement frequency, supplier category, and procurement category, companies can develop a series of targeted management strategies.

[0115] First, formulate a procurement strategy management plan based on the characteristics of the target purchasing group. For example, if the target group shows high-frequency purchases and stable supplier relationships, the company can adopt a bulk purchase or regular purchase strategy. This strategy can not only improve procurement efficiency, but also reduce procurement costs through economies of scale. At the same time, for groups with large procurement amounts, a phased procurement strategy can be adopted to disperse procurement risks.

[0116] Next, companies need to develop a supplier management plan. By analyzing supplier categories and partnerships, companies can identify high-quality suppliers and develop long-term cooperation agreements to establish stable supply chain relationships. For new suppliers, companies should set up strict evaluation and review mechanisms to ensure supplier reliability and product quality. This plan helps companies maintain good supplier relationships and ensure the stability and reliability of the supply chain.

[0117] In terms of procurement planning and execution, companies should formulate detailed procurement plans based on the needs of the target procurement groups. This includes clarifying procurement time nodes, quantities and categories to ensure the feasibility and controllability of procurement plans. Through refined procurement plans, companies can better manage the procurement process and ensure timely delivery and reasonable maintenance of inventory.

[0118] At the same time, companies need to develop inventory management plans. According to the purchase frequency and demand of the target group, companies can set reasonable inventory levels to avoid excessive or low inventory. High-frequency purchasing groups can adopt real-time inventory management strategies to ensure high inventory turnover and low inventory costs. Reasonable inventory management can improve the efficiency of capital use of enterprises and reduce inventory risks.

[0119] In terms of cost control and optimization, enterprises should formulate detailed cost control and optimization plans. By analyzing the cost structure of the target purchasing group, enterprises can identify the key points of cost control. For example, through centralized procurement or negotiation with suppliers, enterprises can obtain more favorable procurement prices. In addition, optimizing the procurement process and reducing unnecessary intermediate links can also help reduce overall procurement costs.

[0120] Finally, companies need to develop risk management plans. Risks in the procurement process include supplier risk, market price fluctuation risk, and delivery risk. Companies should establish a comprehensive risk management mechanism to prevent and respond to potential procurement risks through supplier evaluation, contract management, and market monitoring. Through effective risk management, companies can ensure the smooth progress of the procurement process and reduce cost increases and supply chain disruptions caused by risks.

[0121] The preset management plan is sent to the management terminal of the target enterprise so that the target enterprise can perform procurement management according to the preset management plan.

[0122] First, the preset management plan is converted into specific execution instructions in a systematic way. These instructions cover multiple aspects such as procurement strategy management, supplier management, procurement planning and execution, inventory management, cost control and optimization, and risk management. For example, for the procurement strategy management plan, the system can generate specific purchase orders, time schedules, and quantity planning; for the supplier management plan, it can formulate supplier evaluation standards, contract terms, and cooperation agreements.

[0123] Next, these execution instructions are sent to the management terminal of the target enterprise through the enterprise resource planning (ERP) system or other management information system. These systems can automatically receive and process the instructions of the preset management plan, ensuring that each department and relevant personnel can obtain and understand these instructions in a timely manner. For example, the procurement department can receive specific procurement plans and order information, the supply chain management department can obtain detailed requirements for supplier evaluation and cooperation agreements, and the finance department can understand the specific measures of cost control and optimization plans.

[0124] Through the internal management information system of the enterprise, the preset management plan can be efficiently transmitted and executed among various relevant departments. These systems usually include automated processes and monitoring functions, which can track the execution of various instructions and provide timely feedback on problems and anomalies that occur during the execution process. For example, the system can monitor the execution progress of purchase orders, the delivery status of suppliers, changes in inventory levels, and the effectiveness of cost control. If any deviation from the preset management plan is found, the system can automatically issue an alarm and notify relevant personnel to make adjustments and optimizations.

[0125] Reference Figure 4 , which shows a schematic diagram of the structure of an enterprise intelligent procurement management system based on big data provided by an embodiment of the present application, the system includes: an acquisition module 1, a preprocessing module 2, a feature extraction module 3, a clustering module 4 and a procurement group determination module 5; Acquisition module 1, used to collect historical procurement data from various data sources related to enterprise procurement; Preprocessing module 2, used to preprocess each historical purchase data to obtain the preprocessed historical purchase data corresponding to each data source; Feature extraction module 3, used to extract key procurement data features corresponding to each data source from each pre-processed historical procurement data; the key procurement data features include numerical type features and non-numerical type features; Clustering module 4, used to input each numerical type feature and each non-numerical type feature into a preset clustering analysis model, perform clustering analysis on the purchase data corresponding to each data source, and obtain multiple preset purchase groups; The acquisition module 1 is also used to obtain the current procurement data of the target enterprise; The preprocessing module 2 is also used to preprocess the current purchase data to obtain the preprocessed current purchase data; The purchasing group determination module 5 is used to determine the target purchasing group corresponding to the target enterprise from multiple preset purchasing groups based on the pre-processed current purchasing data.

[0126] In a possible implementation, the preprocessing module 2 is further used to pre-identify each historical purchase data, and determine the data type and preprocessing type corresponding to each historical purchase data; the data type includes a numerical type and a non-numerical type; the preprocessing type includes missing value processing and error value processing; The preprocessing module 2 is further used to perform preliminary processing on the historical purchase data of each numerical type based on missing value processing and / or error value processing to obtain a plurality of preprocessed historical purchase data of numerical types; The preprocessing module 2 is also used to perform preliminary processing on each non-numeric type of historical purchase data based on missing value processing and / or error value processing to obtain multiple preprocessed non-numeric type of historical purchase data.

[0127] In a possible implementation, the clustering module 4 is further used to perform standardization processing on each first statistical feature, each first time feature, each first aggregation feature, and each first derivative feature to obtain a plurality of second statistical features, a plurality of second time features, a plurality of second aggregation features, and a plurality of second derivative features; The clustering module 4 is further used to convert each first classification feature into a binary vector using one-hot encoding to obtain multiple second classification features; The clustering module 4 is further used to represent each first text feature in a corresponding numerical form to obtain a plurality of second text features; The clustering module 4 is further used to integrate the plurality of second statistical features, the plurality of second time features, the plurality of second aggregation features, the plurality of second derivative features, the plurality of second classification features, and the plurality of second text features into a unified data frame, wherein the data frame includes a plurality of data points, each row of the data frame represents a purchase record, and each column of the data frame represents a feature; The clustering module 4 is also used to determine the clustering algorithm and clustering parameters; The clustering module 4 is also used to input the data frame into the clustering algorithm and perform training based on the clustering parameters to obtain the cluster label corresponding to each data point; The clustering module 4 is further used to obtain multiple preset purchasing groups based on each cluster label.

[0128] In a possible implementation, the purchasing group determination module 5 is further used to extract a third statistical feature, a third time feature, a third aggregation feature, a third derivative feature, a third classification feature, and a third text feature from the preprocessed current purchasing data; The purchasing group determination module 5 is further used to calculate the statistical feature similarity between the third statistical feature and each second statistical feature, and calculate the time feature similarity between the third time feature and each second time feature, and calculate the aggregation feature similarity between the third aggregation feature and each second aggregation feature, and calculate the derivative feature similarity between the third derivative feature and each second derivative feature, and calculate the classification feature similarity between the third classification feature and each second classification feature, and calculate the text feature similarity between the third text feature and each second text feature; The purchasing group determination module 5 is further used to perform weighted summation of the statistical feature similarity, time feature similarity, aggregation feature similarity, derivative feature similarity, classification feature similarity and text feature similarity corresponding to each preset purchasing group to obtain a total similarity score corresponding to each preset purchasing group; The purchasing group determination module 5 is also used to select a preset purchasing group with the highest total similarity score as the target purchasing group.

[0129] In a possible implementation, the purchasing group determination module 5 is further used to calculate the statistical feature similarity, the time feature similarity and the aggregate feature similarity based on the Euclidean distance or cosine similarity calculation method; The purchasing group determination module 5 is further used to calculate the classification feature similarity based on the Jaccard similarity or cosine similarity calculation method; The purchasing group determination module 5 is also used to calculate the text feature similarity based on the TF-IDF cosine similarity calculation method.

[0130] In a possible implementation, the system further includes a management module 6; the management module 6 is used to determine a preset management plan corresponding to the target purchasing group; wherein the preset management plan includes a purchasing strategy management plan, a supplier management plan, a purchasing plan and execution plan, an inventory management plan, a cost control and optimization plan, and a risk management plan; The management module 6 is also used to send the preset management plan to the management terminal of the target enterprise, so that the target enterprise can perform procurement management according to the preset management plan.

[0131] It should be noted that: when the device provided in the above embodiment realizes its function, only the division of the above functional modules is used as an example. In actual application, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0132] The present application also discloses an electronic device. Figure 5 , Figure 5 The electronic device 500 may include: at least one processor 501 , at least one network interface 504 , a user interface 503 , a memory 505 , and at least one communication bus 502 .

[0133] The communication bus 502 is used to realize the connection and communication between these components.

[0134] The user interface 503 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 503 may also include a standard wired interface and a wireless interface.

[0135] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0136] Among them, the processor 501 may include one or more processing cores. The processor 501 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 505, and calling data stored in the memory 505. Optionally, the processor 501 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 501 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processor (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 501, and it can be implemented separately through a chip.

[0137] Among them, the memory 505 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 505 includes a non-transitory computer-readable storage medium. The memory 505 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 505 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned method embodiments, etc.; the data storage area may store data involved in the above-mentioned method embodiments, etc. The memory 505 may optionally be at least one storage device located away from the aforementioned processor 501. Refer to Figure 5 , the memory 505 as a computer-readable storage medium may include an operating system, a network communication module, a user interface module and an application program.

[0138] exist Figure 5 In the electronic device 500 shown, the user interface 503 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 501 can be used to call an application stored in the memory 505, and when executed by one or more processors 501, the electronic device 500 executes one or more methods in the above-mentioned embodiments. It should be noted that for the aforementioned method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application.

[0139] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0140] In the several embodiments provided in this application, it should be understood that the disclosed devices can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0141] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0142] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0143] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a memory and includes several instructions for a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0144] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0145] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. An enterprise intelligent procurement management method based on big data, characterized in that: The method comprises: Collect historical procurement data from various data sources related to corporate procurement; Preprocessing each of the historical purchase data to obtain the preprocessed historical purchase data corresponding to each of the data sources; Extracting key procurement data features corresponding to each of the data sources from each of the preprocessed historical procurement data; the key procurement data features include numerical type features and non-numerical type features; Inputting each of the numerical type features and each of the non-numerical type features into a preset cluster analysis model, performing cluster analysis on the purchase data corresponding to each of the data sources, and obtaining a plurality of preset purchase groups; Obtain current purchasing data of target enterprises; Preprocessing the current purchase data to obtain preprocessed current purchase data; Based on the preprocessed current procurement data, a target procurement group corresponding to the target enterprise is determined from a plurality of preset procurement groups.

2. The method according to claim 1, characterized in that The preprocessed historical purchase data includes preprocessed historical purchase data of numerical type and preprocessed historical purchase data of non-numerical type; the preprocessing of each of the historical purchase data to obtain the preprocessed historical purchase data corresponding to each of the data sources specifically includes: Pre-identify each of the historical purchase data to determine the data type and pre-processing type corresponding to each of the historical purchase data; the data type includes a numerical type and a non-numerical type; the pre-processing type includes missing value processing and error value processing; Based on the missing value processing and / or the error value processing, respectively, the historical purchase data of each numerical type are preliminarily processed to obtain a plurality of preprocessed historical purchase data of the numerical type; Based on the missing value processing and / or the error value processing, each non-numeric type of historical purchase data is preliminarily processed to obtain a plurality of pre-processed non-numeric type of historical purchase data.

3. The method according to claim 1, characterized in that The numerical type features include a first statistical feature, a first time feature, a first aggregation feature, and a first derivative feature; the non-numerical type features include a first classification feature and a first text feature.

4. The method according to claim 3, characterized in that: The step of inputting each of the numerical type features and each of the non-numerical type features into a preset cluster analysis model, performing cluster analysis on the purchase data corresponding to each of the data sources, and obtaining a plurality of preset purchase groups specifically includes: Performing standardization processing on each of the first statistical features, each of the first time features, each of the first aggregate features, and each of the first derived features to obtain a plurality of second statistical features, a plurality of second time features, a plurality of second aggregate features, and a plurality of second derived features; Using one-hot encoding to convert each of the first classification features into a binary vector to obtain a plurality of second classification features; Representing each of the first text features in a corresponding numerical form to obtain a plurality of second text features; Integrate a plurality of the second statistical features, a plurality of the second time features, a plurality of the second aggregate features, a plurality of the second derivative features, a plurality of the second classification features, and a plurality of the second text features into a unified data frame, wherein the data frame includes a plurality of data points, each row of the data frame represents a purchase record, and each column of the data frame represents a feature; Determine the clustering algorithm and clustering parameters; Inputting the data frame into the clustering algorithm and performing training based on the clustering parameters to obtain a cluster label corresponding to each of the data points; Based on each of the cluster labels, a plurality of the preset purchasing groups are obtained.

5. The method according to claim 1, characterized in that The determining, based on the pre-processed current procurement data, a target procurement group corresponding to the target enterprise from a plurality of preset procurement groups specifically includes: extracting a third statistical feature, a third time feature, a third aggregation feature, a third derivative feature, a third classification feature, and a third text feature from the preprocessed current purchase data; Calculating the statistical feature similarity between the third statistical feature and each of the second statistical features, calculating the time feature similarity between the third time feature and each of the second time features, calculating the aggregate feature similarity between the third aggregate feature and each of the second aggregate features, calculating the derivative feature similarity between the third derivative feature and each of the second derivative features, calculating the classification feature similarity between the third classification feature and each of the second classification features, and calculating the text feature similarity between the third text feature and each of the second text features; The statistical feature similarity, the time feature similarity, the aggregate feature similarity, the derivative feature similarity, the classification feature similarity and the text feature similarity corresponding to each of the preset purchasing groups are weighted and summed to obtain a total similarity score corresponding to each of the preset purchasing groups; The preset purchasing group with the highest total similarity score is selected as the target purchasing group.

6. The method according to claim 5, characterized in that The method further comprises: Based on the Euclidean distance or cosine similarity calculation method, the statistical feature similarity, the time feature similarity and the aggregate feature similarity are calculated; Calculating the classification feature similarity based on the Jaccard similarity or cosine similarity calculation method; Based on the TF-IDF cosine similarity calculation method, the text feature similarity is calculated.

7. The method according to claim 1, characterized in that After determining the target purchasing group corresponding to the target enterprise from the plurality of preset purchasing groups based on the preprocessed current purchasing data, the method further includes: Determine the preset management plan corresponding to the target purchasing group; wherein the preset management plan includes a purchasing strategy management plan, a supplier management plan, a purchasing plan and execution plan, an inventory management plan, a cost control and optimization plan, and a risk management plan; The preset management plan is sent to the management terminal of the target enterprise, so that the target enterprise performs procurement management according to the preset management plan.

8. An enterprise intelligent procurement management system based on big data, characterized in that: The system includes: an acquisition module, a preprocessing module, a feature extraction module, a clustering module and a purchasing group determination module; The acquisition module is used to collect historical procurement data from various data sources related to enterprise procurement; The preprocessing module is used to preprocess each of the historical purchase data to obtain the preprocessed historical purchase data corresponding to each of the data sources; The feature extraction module is used to extract key purchase data features corresponding to each of the data sources from each of the preprocessed historical purchase data; the key purchase data features include numerical type features and non-numerical type features; The clustering module is used to input each of the numerical type features and each of the non-numerical type features into a preset clustering analysis model, perform clustering analysis on the purchase data corresponding to each of the data sources, and obtain multiple preset purchase groups; The acquisition module is also used to acquire the current procurement data of the target enterprise; The preprocessing module is further used to preprocess the current purchase data to obtain the preprocessed current purchase data; The purchasing group determination module is used to determine the target purchasing group corresponding to the target enterprise from a plurality of preset purchasing groups based on the pre-processed current purchasing data.

9. An electronic device, characterized in that: The electronic device (500) comprises a processor (501), a memory (505), a user interface (503) and a network interface (504), wherein the memory (505) is used to store instructions, the user interface (503) and the network interface (504) are used to communicate with other devices, and the processor (501) is used to execute the instructions stored in the memory (505) so that the electronic device (500) executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed, the method according to any one of claims 1 to 7 is executed.

Images