Automatic purchasing system and method based on artificial intelligence

Abstract

The invention provides an automatic procurement system and method based on artificial intelligence, and the method comprises the steps: carrying out the preprocessing of a multi-modal data set related to procurement, and constructing an automatic procurement model based on multi-modal learning; training the automatic purchasing model by using the training set to obtain a trained automatic purchasing model; performing accuracy and efficiency evaluation on the trained automatic purchase model through the verification set, and adjusting parameters of the automatic purchase model to obtain a final automatic purchase model; and acquiring real-time purchasing demand data, generating an optimal purchasing scheme by using the final automatic purchasing model, and executing a purchasing task, so that intelligent decision-making of the purchasing scheme can be realized, full-process automatic purchasing can be realized, errors caused by manual data collection and analysis can be avoided, and the purchasing accuracy and efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of supply chain technology, and in particular to an automatic procurement system and method based on artificial intelligence. Background Art

[0002] The traditional procurement process is plagued by complex procedures, high labor costs, low efficiency, delayed response times, and subjective supplier selection. Procurement personnel spend a significant amount of time screening suppliers, comparing prices, and negotiating contracts. This approach is not only cumbersome, with long procurement cycles and high labor costs, but also relies heavily on manual calculations and operations, making it prone to human errors such as data entry errors and miscalculations, resulting in low procurement accuracy and efficiency.

[0003] Semi-automated procurement technologies enable online submission of purchase applications and automated approval processes. However, limitations remain. For example, data collection and analysis still require significant manual intervention, and data from different systems is difficult to effectively integrate, hindering comprehensive intelligent decision-making.

[0004] When faced with massive amounts of market information and complex procurement needs, existing procurement technologies are unable to make optimal procurement decisions quickly and accurately. Summary of the Invention

[0005] In order to solve the above technical problems, this application provides an automatic procurement system and method based on artificial intelligence.

[0006] In a first aspect of the present application, an artificial intelligence-based automatic procurement method is provided, the method comprising the following steps: Step S1, obtaining a procurement-related multimodal dataset; Step S2: preprocessing the multimodal dataset to obtain a preprocessed multimodal dataset, and dividing the preprocessed multimodal dataset into a training set and a validation set according to a set ratio; Step S3, building an automatic procurement model based on multimodal learning; Step S4: training the automatic procurement model using the training set to obtain a trained automatic procurement model; Step S5: Evaluate the accuracy and efficiency of the trained automatic procurement model using the validation set, and adjust the parameters of the automatic procurement model to obtain the final automatic procurement model; Step S6: Acquire real-time procurement demand data, use the final automatic procurement model to generate a procurement plan, and execute the procurement task.

[0007] In some embodiments of the present application, step S2 includes: Step S21, performing data cleaning on the multimodal dataset to deal with existing errors, duplicate values ​​or outliers; Step S22, performing data normalization on the cleaned multimodal dataset to obtain a normalized multimodal dataset; Step S23, performing data enhancement on the standardized multimodal dataset to obtain an enhanced multimodal dataset; In step S24, the enhanced multimodal dataset is divided into a training set and a validation set according to a set ratio.

[0008] In some embodiments of the present application, in step S22, For numerical data, the Z-score standardization method is used to convert it into a standard normal distribution with a mean of 0 and a standard deviation of 1, that is,

[0009] in, is the original data, is the mean of the data, is the standard deviation of the data; For text data, the TF-IDF algorithm is used to extract keywords and convert them into vector form for subsequent semantic embedding processing.

[0010] In some embodiments of the present application, step S3 includes: Step S31, inputting the standardized multimodal data set into a data fusion module for feature extraction to obtain a feature vector; Step S32, performing feature fusion on the feature vectors using a weighted fusion strategy; Step S33: Building an automatic procurement model based on multimodal learning.

[0011] In some embodiments of the present application, step S31 specifically includes: For the For commodities, we use natural language processing technology to extract semantic features from text data, and extract transaction frequency, average transaction amount, transaction satisfaction score, etc. from the supplier's historical transaction records as cooperation features. ; Use sentiment analysis algorithms to extract the semantic representation of keywords from product quality evaluations in order to obtain product quality characteristics ; Get the real-time market price of each supplier's goods as the price feature ; Get the average logistics transportation time of each supplier as the service characteristic ; Through the long short-term memory network, the historical market price fluctuation data and the characteristics of seasonal factors affecting the market price fluctuation data are extracted as price fluctuation features .

[0012] In some embodiments of the present application, feature fusion is performed on feature vectors based on a weighted fusion strategy; The cooperative features , quality characteristics , price characteristics , Service Features , price fluctuation characteristics Perform fusion and set the fused features to ,but:

[0013] in, 、 、 、 、 is the weight coefficient. The weights of different characteristics can be adjusted accordingly according to the performance of the product or the adjustment of the procurement strategy.

[0014] In some embodiments of the present application, step S33 specifically includes: Fusion features , procurement demand as input, procurement strategy as output, and construct a multimodal learning model based on multimodal learning, as follows:

[0015] in, Multilayer perceptron representing a multimodal learning model for decoding fused features And output procurement strategy .

[0016] In a second aspect of the present application, an artificial intelligence-based automatic procurement system is provided, comprising a data acquisition module, a data preprocessing module, a data fusion module, an intelligent decision-making module, and an execution feedback module; The data acquisition module is used to obtain procurement-related multimodal data sets; The data preprocessing module is used to preprocess the multimodal dataset and divide the preprocessed multimodal dataset into a training set and a validation set according to a set ratio; The data fusion module is used to extract features from the pre-processed multimodal data, obtain feature vectors and perform feature fusion on the feature vectors; The intelligent decision-making module is used to build an automatic procurement model and generate a procurement strategy; The execution feedback module is used to receive the procurement strategy and monitor and adjust the procurement strategy.

[0017] In some embodiments of the present application, the data preprocessing module includes a data cleaning unit, a data normalization unit, a data enhancement unit, and a data partitioning unit; The data cleaning unit is used to perform data cleaning on the multimodal data set; The data standardization unit is used to perform data standardization on the multimodal data set that has completed data cleaning; The data enhancement unit is used to perform data enhancement on the standardized multimodal dataset; The data partitioning unit is used to divide the enhanced multimodal dataset into a training set and a validation set according to a set ratio.

[0018] Compared with the prior art, the present invention has the following advantages and beneficial effects: in the artificial intelligence-based automatic procurement method of the present application, an automatic procurement model based on multimodal learning is constructed by preprocessing the procurement-related multimodal data set; and the automatic procurement model is trained using a training set to obtain a trained automatic procurement model; the accuracy and efficiency of the trained automatic procurement model are evaluated using a validation set, and the parameters of the automatic procurement model are adjusted to obtain a final automatic procurement model; real-time procurement demand data is obtained, and the final automatic procurement model is used to generate the optimal procurement plan and execute the procurement task. In this way, intelligent decision-making of the procurement plan can be achieved, and automated procurement of the entire process can be achieved, errors caused by manual data collection and analysis can be avoided, and the accuracy and efficiency of procurement can be improved.

[0019] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which constitute part of this document, are intended to provide a further understanding of this document. The exemplary embodiments and descriptions herein are intended to explain this document and do not constitute an improper limitation on this document. In the accompanying drawings: Figure 1 is a flow chart of an automatic procurement method based on artificial intelligence provided by an exemplary embodiment of the present application; Figure 2 This is a principle block diagram of an artificial intelligence-based automatic procurement system provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. It should be noted that, in the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined with each other in any way.

[0022] The traditional procurement process is plagued by complex procedures, high labor costs, low efficiency, delayed response times, and subjective supplier selection. Procurement personnel spend a significant amount of time screening suppliers, comparing prices, and negotiating contracts. This approach is not only cumbersome, with long procurement cycles and high labor costs, but also relies heavily on manual calculations and operations, making it prone to human errors such as data entry errors and miscalculations, resulting in low procurement accuracy and efficiency.

[0023] Semi-automated procurement technologies enable online submission of purchase applications and automated approval processes. However, limitations remain. For example, data collection and analysis still require significant manual intervention, and data from different systems is difficult to effectively integrate, hindering comprehensive intelligent decision-making.

[0024] When faced with massive amounts of market information and complex procurement needs, existing procurement technologies are unable to make optimal procurement decisions quickly and accurately.

[0025] Based on this, an exemplary embodiment of the present application provides an automatic procurement method based on artificial intelligence, in which an automatic procurement model based on multimodal learning is constructed by preprocessing procurement-related multimodal data sets; and the automatic procurement model is trained using a training set to obtain a trained automatic procurement model; the accuracy and efficiency of the trained automatic procurement model are evaluated using a validation set, and the parameters of the automatic procurement model are adjusted to obtain a final automatic procurement model; real-time procurement demand data is obtained, and the final automatic procurement model is used to generate the optimal procurement plan and execute procurement tasks. In this way, intelligent decision-making of procurement plans can be achieved, and automated procurement of the entire process can be achieved, errors caused by manual data collection and analysis can be avoided, and the accuracy and efficiency of procurement can be improved.

[0026] Example 1: An exemplary embodiment of the present application provides an automatic procurement method based on artificial intelligence, such as Figure 1 As shown, the method includes the following steps: In step S1, the data acquisition module acquires a multimodal dataset related to procurement. This dataset includes historical supplier transaction records, product quality ratings, real-time market prices, logistics transportation times, procurement requirements, and historical product price fluctuations. For example, if the data acquisition module is connected to an ERP system or warehouse management system, the supplier's historical transaction records can be accessed through the ERP system, reflecting information such as supplier credibility, supply quality, and price stability. Web crawler technology can simulate human online search and web browsing behaviors, automatically acquiring real-time market prices and quality ratings from major e-commerce platforms and supplier websites. For example, it can automatically access the websites of 100 commonly used suppliers every hour to update product market prices in real time to ensure data timeliness. Logistics transportation times can be obtained from third-party logistics platforms. Procurement requirements include key information such as the name, quantity, budget, and arrival time of the purchased product. The procurement requirements determine the name of the purchased product and other information, and then acquire information such as the real-time market price, product quality rating, and historical supplier transaction records. Historical product price fluctuation data includes historical market price fluctuation data and market price fluctuation data affected by seasonal factors; web crawler technology can be used to collect historical market price fluctuation data and market price fluctuation data due to seasonal factors.

[0027] Step S2: preprocess the multimodal dataset to obtain a preprocessed multimodal dataset, and divide the preprocessed multimodal dataset into a training set and a validation set according to a set ratio.

[0028] The preprocessing of multimodal datasets includes cleaning, standardization, and data enhancement.

[0029] The specific steps include: Step S21 is to perform data cleaning on the multimodal dataset, which mainly includes field alignment and outlier correction to deal with existing errors, duplicate values, or outliers. The purpose of field alignment is to map data fields from different sources to a unified standard field format. For example, the "transaction amount" field in a supplier's historical transaction records may use different currency units, the "transaction time" field may use different time formats, and the "price" field in real-time market price data may also use different formats. To ensure data consistency, all currency units need to be unified into a standard currency, and all time fields need to be converted to a unified time format.

[0030] Outlier correction is the application of statistical methods to address errors or outliers in data. For example, if a supplier's historical transaction records contain price data that significantly deviates from the normal range, we can calculate the mean and standard deviation of that supplier's historical transaction prices and deem any data outside the range of three standard deviations as outliers, thereby applying corrections.

[0031] Step S22: After data cleaning is completed, the multimodal dataset is normalized to eliminate dimensional differences and obtain a normalized multimodal dataset.

[0032] For numerical data, such as historical transaction prices of suppliers, real-time market prices of commodities, etc., the Z-score standardization method is used to convert them into a standard normal distribution with a mean of 0 and a standard deviation of 1, that is,

[0033] in, is the original data, is the mean of the data, is the standard deviation of the data. By standardizing numerical data, data from different sources can be compared and analyzed using the same scale. For textual data, the TF-IDF algorithm is used to extract keywords and convert them into vector form for subsequent semantic embedding.

[0034] Step S23, performing data enhancement on the standardized multimodal dataset to obtain an enhanced multimodal dataset; In step S24, the enhanced multimodal dataset is divided into a training set and a validation set according to a set ratio. For example, 70% of the data is divided into a training set and 30% of the data is divided into a validation set.

[0035] Step S3: constructing an automatic procurement model based on multimodal learning; specifically, the following steps are included: In step S31 , the standardized multimodal data set is input into a data fusion module for feature extraction to obtain a feature vector.

[0036] Specifically, for example, For commodities, we use natural language processing technology to extract semantic features from text data, and extract transaction frequency, average transaction amount, transaction satisfaction score, etc. from the supplier's historical transaction records as cooperation features. ; Use sentiment analysis algorithms to extract the semantic representation of keywords from product quality evaluations in order to obtain product quality characteristics For example, when purchasing silicon carbide rods, we can obtain 100 product quality reviews from each supplier, extract keywords, and classify the keywords into three types: positive, neutral, and negative. We calculate the number and score of the three keywords respectively, and obtain the comprehensive quality score of the product as the quality feature after weighting. ; Get the real-time market price of each supplier's goods as a price feature ; Get the average logistics transportation time of each supplier as the service characteristic ; The long short-term memory network is used to extract historical market price fluctuation data and the characteristics of seasonal factors affecting market price fluctuation data as price fluctuation features. ; Exemplarily, a multi-layer LSTM network structure is adopted, which includes multiple LSTM layers, each layer includes several memory units, and the memory units realize the memory and forgetting functions of time series data through a gating mechanism. Each memory unit contains three parts: input gate, forget gate and output gate. LSTM can capture the time dependency of data over a long time span while avoiding the gradient vanishing problem. Historical market price fluctuation data is input into the first layer of LSTM, and preliminary time series features such as price fluctuation trends and amplitude changes are extracted from the original input data; market price fluctuation data affected by seasonal factors is input into the second layer of LSTM, and seasonal impact data such as cyclical fluctuations and seasonal patterns are extracted; historical market price fluctuation data and market price fluctuation data affected by seasonal factors are spliced ​​into a comprehensive feature vector in the feature fusion layer, and price fluctuation features are output in the output layer. The layer-by-layer approach enables the network to gradually abstract the key features in the data.

[0037] Preferably, batch normalization is added after each LSTM layer to reduce internal covariate shift. After feature learning, the LSTM network automatically extracts deeper time series features and cyclical features. Specifically, historical market price fluctuation data and seasonally influenced market price fluctuation data are fed into the multi-layer LSTM network for forward propagation. Within each LSTM layer, memory cells dynamically adjust their gating parameters based on the state of the input data, capturing features at different time scales. For example, for historical market price fluctuation data, LSTM can identify short-term price fluctuation patterns and long-term trend changes; for seasonally influenced market price fluctuation data, LSTM can detect cyclical features such as holiday effects and seasonal demand fluctuations. Features are extracted layer by layer and passed to higher layers of the network, such as the feature fusion layer, to form a comprehensive feature. To further optimize feature extraction, an attention mechanism can be added to the top layer of the network. By calculating attention weights at each time step, it can improve feature representativeness.

[0038] Step S32, performing feature fusion on the feature vectors using a weighted fusion strategy; The cooperative features , quality characteristics , price characteristics , Service Features , price fluctuation characteristics Perform fusion and set the fused features to , In this application, weighted fusion is adopted, then:

[0039] in, 、 、 、 、 is the weight coefficient, and the weights of different features can be adjusted accordingly according to the performance of the product or the adjustment of the procurement strategy. For example, when the purchased product is silicon carbide rod, the price feature can be set according to the procurement requirements. The weight coefficient 5, quality characteristics The weight coefficient 5. Service Features The weight coefficient 2. Cooperation characteristics The weight coefficient 3. Price fluctuation characteristics The weight coefficient is 3, to obtain fusion features .

[0040] Step S33: constructing an automatic procurement model based on multimodal learning; Fusion features , procurement demand as input, procurement strategy as output, and construct a multimodal learning model based on multimodal learning, as follows:

[0041] in, Multilayer perceptron representing a multimodal learning model for decoding fused features And output procurement strategy . Procurement Strategy It should at least include confirmation of the supplier to be purchased, the purchase time, and the quantity of goods to be purchased.

[0042] Step S4: training the automatic procurement model using the training set to obtain a trained automatic procurement model; Step S5: The accuracy and efficiency of the trained automatic procurement model are evaluated using the validation set, and the parameters of the automatic procurement model are adjusted to obtain the final automatic procurement model.

[0043] Step S6: Real-time procurement demand data is acquired, and the final automated procurement model is used to generate the optimal procurement plan and execute the procurement task. After the procurement plan is confirmed, automated process robotics technology is used to automatically complete operations such as purchase order issuance, electronic contract signing, and payment processing, ensuring efficient execution of the procurement process.

[0044] Step S7: monitor and adjust the generated procurement plan in real time.

[0045] Example 2: An exemplary embodiment of the present application provides an automatic procurement system based on artificial intelligence, such as Figure 2 As shown in the figure, the system includes a data acquisition module, a data preprocessing module, a data fusion module, an intelligent decision-making module and an execution feedback module; among them, the multimodal data set includes the supplier's historical transaction records, product quality evaluation, real-time market price of the product, logistics transportation time, procurement demand, historical product price fluctuation data, etc.

[0046] The data preprocessing module is used to preprocess the multimodal dataset. The preprocessing may include cleaning, standardization, and data enhancement. The preprocessed multimodal dataset is then divided into a training set and a validation set according to a set ratio. The data fusion module is used to extract features from the preprocessed multimodal data, obtain feature vectors and perform feature fusion on the feature vectors; The intelligent decision-making module is used to build automatic procurement models and generate procurement strategies; The execution feedback module is used to receive, monitor and adjust the procurement strategy; Among them, the data preprocessing module includes a data cleaning unit, a data standardization unit, a data enhancement unit and a data partitioning unit; The data cleaning unit is used to clean multimodal data sets, such as field alignment and outlier correction; field alignment is to map data fields from different sources to a unified standard field format; outlier correction is to use statistical methods to correct data with outliers.

[0047] The data standardization unit is used to standardize the multimodal dataset that has completed data cleaning, such as unifying the currency unit into a standard currency and converting the time field into a unified time format.

[0048] The data enhancement unit is used to perform data enhancement on the standardized multimodal dataset; The data partitioning unit is used to divide the enhanced multimodal dataset into a training set and a validation set according to a set ratio.

[0049] The data fusion module includes a semantic embedding unit, a structural embedding unit, a relational embedding unit, and a feature fusion unit; the semantic embedding unit is used to represent the semantic information in text data; the structural embedding unit is used to represent the hierarchy and association relationship between data; the relational embedding unit is used to represent the interaction relationship between entities; the feature fusion unit is used to perform weighted fusion on the results of semantic embedding, structural embedding, and relational embedding to obtain regularized multimodal features.

[0050] Preferably, the system may also include a human-computer interaction module that utilizes speech recognition and natural language generation technologies to enable natural interaction with purchasing personnel. For example, after a purchase order is created, the purchasing personnel can query the purchase progress through the human-computer interaction module. Based on the received voice commands, the system queries the corresponding purchase order and announces or displays the detailed purchase order information via voice or text, allowing the purchasing personnel to be informed of the execution status of the purchase order at any time.

[0051] The execution feedback module also generates procurement decision execution reports at preset intervals. These reports include product name, price, supplier information, order creation time, delivery time, payment status, and other information, allowing procurement personnel to access and review them. When anomalies occur in a purchase order, such as unexpected market price fluctuations, supplier default, or significant changes in inventory status, the system identifies the anomaly and, based on the anomaly, current product inventory, and production order information, reassesses procurement demand using the automated procurement model, dynamically adjusts procurement plans, and regenerates new purchase orders.

[0052] For example, when there is a significant fluctuation in market prices, for example, when the current market price is lower than 90% of the order price, the procurement demand is re-evaluated.

[0053] In this application, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that an article or device comprising a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such article or device. In the absence of further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the article or device comprising the element.

[0054] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0055] Obviously, those skilled in the art may make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if such changes and modifications of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such changes and modifications.

Claims

1. An automatic procurement method based on artificial intelligence, characterized in that: The method comprises the following steps: Step S1, obtaining a procurement-related multimodal dataset; Step S2: preprocessing the multimodal dataset to obtain a preprocessed multimodal dataset, and dividing the preprocessed multimodal dataset into a training set and a validation set according to a set ratio; Step S3, building an automatic procurement model based on multimodal learning; Step S4: training the automatic procurement model using the training set to obtain a trained automatic procurement model; Step S5: Evaluate the accuracy and efficiency of the trained automatic procurement model using the validation set, and adjust the parameters of the automatic procurement model to obtain the final automatic procurement model; Step S6: Acquire real-time procurement demand data, use the final automatic procurement model to generate a procurement plan, and execute the procurement task.

2. The automatic procurement method based on artificial intelligence according to claim 1, characterized in that: The step S2 comprises: Step S21, performing data cleaning on the multimodal dataset to deal with existing errors, duplicate values ​​or outliers; Step S22, performing data normalization on the cleaned multimodal dataset to obtain a normalized multimodal dataset; Step S23, performing data enhancement on the standardized multimodal dataset to obtain an enhanced multimodal dataset; In step S24, the enhanced multimodal dataset is divided into a training set and a validation set according to a set ratio.

3. The automatic procurement method based on artificial intelligence according to claim 2, characterized in that: In the step S22, For numerical data, the Z-score standardization method is used to convert it into a standard normal distribution with a mean of 0 and a standard deviation of 1, that is, in, is the original data, is the mean of the data, is the standard deviation of the data; For text data, the TF-IDF algorithm is used to extract keywords and convert them into vector form for subsequent semantic embedding processing.

4. The automatic procurement method based on artificial intelligence according to claim 1, characterized in that: The step S3 comprises: Step S31, inputting the standardized multimodal data set into a data fusion module for feature extraction to obtain a feature vector; Step S32, performing feature fusion on the feature vectors using a weighted fusion strategy; Step S33: Building an automatic procurement model based on multimodal learning.

5. The automatic procurement method based on artificial intelligence according to claim 4 is characterized in that: The step S31 specifically includes: For the For commodities, we use natural language processing technology to extract semantic features from text data, and extract transaction frequency, average transaction amount, transaction satisfaction score, etc. from the supplier's historical transaction records as cooperation features. ; Use sentiment analysis algorithms to extract the semantic representation of keywords from product quality evaluations in order to obtain product quality characteristics ; Get the real-time market price of each supplier's goods as the price feature ; Get the average logistics transportation time of each supplier as the service characteristic ; Through the long short-term memory network, the historical market price fluctuation data and the characteristics of seasonal factors affecting the market price fluctuation data are extracted as price fluctuation features .

6. The automatic procurement method based on artificial intelligence according to claim 5, characterized in that: The feature vector is fused based on weighted fusion strategy; The cooperative features , quality characteristics , price characteristics , Service Features , price fluctuation characteristics Perform fusion and set the fused features to ,but: in, 、 、 、 、 is the weight coefficient. The weights of different characteristics can be adjusted accordingly according to the performance of the product or the adjustment of the procurement strategy.

7. The automatic procurement method based on artificial intelligence according to claim 4, characterized in that: The step S33 specifically includes: Fusion features , procurement demand as input, procurement strategy as output, and construct a multimodal learning model based on multimodal learning, as follows: in, Multilayer perceptron representing a multimodal learning model for decoding fused features And output procurement strategy .

8. The automatic procurement method based on artificial intelligence according to claim 1, characterized in that: The following steps are also included: Step S7: monitor and adjust the generated procurement plan in real time.

9. An automatic purchasing system based on artificial intelligence, characterized in that: It includes data acquisition module, data preprocessing module, data fusion module, intelligent decision-making module and execution feedback module; The data acquisition module is used to obtain procurement-related multimodal data sets; The data preprocessing module is used to preprocess the multimodal dataset and divide the preprocessed multimodal dataset into a training set and a validation set according to a set ratio; The data fusion module is used to extract features from the pre-processed multimodal data, obtain feature vectors and perform feature fusion on the feature vectors; The intelligent decision-making module is used to build an automatic procurement model and generate a procurement strategy; The execution feedback module is used to receive the procurement strategy and monitor and adjust the procurement strategy.

10. The automatic purchasing system based on artificial intelligence according to claim 7, characterized in that: The data preprocessing module includes a data cleaning unit, a data standardization unit, a data enhancement unit and a data division unit; The data cleaning unit is used to perform data cleaning on the multimodal data set; The data standardization unit is used to perform data standardization on the multimodal data set that has completed data cleaning; The data enhancement unit is used to perform data enhancement on the standardized multimodal dataset; The data partitioning unit is used to divide the enhanced multimodal dataset into a training set and a validation set according to a set ratio.

Images