A smart park investment recommendation method and system based on machine learning

Abstract

This invention discloses a machine learning-based method and system for investment promotion recommendation in smart industrial parks, relating to the field of smart industrial park management technology. The method includes: collecting multi-source heterogeneous data from enterprises and park infrastructure within the smart park, extracting and embedding features to construct an enterprise feature matrix and a park resource feature matrix; constructing a multi-task prediction model using machine learning algorithms based on the enterprise feature matrix and park resource feature matrix; constructing an investment promotion recommendation evaluation function and introducing an improved slime mold optimization algorithm to optimize the weight vector, resulting in an optimized investment promotion recommendation evaluation function; and generating investment promotion recommendation results using the multi-task prediction model and the optimized investment promotion recommendation evaluation function based on the enterprise feature matrix and park resource feature matrix. This invention effectively solves the problems of low accuracy in virtual-real fusion, lack of adaptability in teaching models, single interaction methods, and single evaluation dimensions by deeply mining data features and adaptively optimizing evaluation weights.

Description

Technical Field

[0001] This invention relates to the field of smart park management technology, and in particular to a smart park investment recommendation method and system based on machine learning. Background Technology

[0002] With the development of urban economies, industrial parks have become important platforms for promoting industrial agglomeration and driving regional economic growth. However, traditional park investment promotion models often rely on manual introductions, offline property viewings, and simple information matching, which suffer from problems such as information asymmetry, low matching efficiency, and high subjectivity. On the one hand, park managers find it difficult to accurately identify high-quality enterprises that match the park's industrial positioning and have high growth potential from massive amounts of enterprise data; on the other hand, enterprises also find it difficult to find the most suitable resources for their development within the complex park environment.

[0003] Most existing investment promotion recommendation systems employ rule-based matching methods (such as matching only industry codes or area requirements) or simple collaborative filtering algorithms, lacking the ability to uncover the deep semantic relationships between enterprises and industrial parks. For example, the potential needs implied in a company's business scope text are often overlooked. Furthermore, investment promotion is a complex multi-objective decision-making process, and existing solutions often focus only on a single dimension of matching (such as whether a company will move in), failing to comprehensively consider multiple evaluation indicators such as the probability of company entry, annual output forecasts, and industrial chain synergy effects. More importantly, in existing evaluation models, the weights of each indicator are usually set by expert experience, lacking adaptability and failing to cope with complex and ever-changing market environments and the differentiated needs of different industrial parks.

[0004] Therefore, there is an urgent need for a smart park investment promotion recommendation method that can deeply mine multi-source heterogeneous data, integrate multi-dimensional indicators, and has adaptive optimization capabilities. Summary of the Invention

[0005] This invention provides a smart park investment recommendation method and system based on machine learning, which solves the problems of low accuracy of virtual-real fusion, lack of adaptability of teaching mode, single interaction method and single evaluation dimension in existing technologies.

[0006] In a first aspect, embodiments of the present invention provide a smart park investment recommendation method based on machine learning, the method comprising: Collect multi-source heterogeneous data of enterprises and park facilities in the smart park, extract and embed features, and construct enterprise feature matrix and park resource feature matrix of the smart park. Based on the enterprise feature matrix and park resource feature matrix of the smart park, a multi-task prediction model is constructed using machine learning algorithms to predict investment recommendation indicators for enterprises and park facilities. Based on the investment recommendation indicators, an investment recommendation evaluation function is constructed, and an improved slime mold optimization algorithm is introduced to optimize the weight vector of the investment recommendation evaluation function, thus obtaining the optimized investment recommendation evaluation function. Based on the enterprise characteristic matrix and park resource characteristic matrix of the smart park, a multi-task prediction model and an optimized investment promotion recommendation evaluation function are used to generate investment promotion recommendation results for enterprises in the smart park.

[0007] The technical solution provided in this application has at least the following beneficial effects: By constructing a multi-task prediction model based on machine learning algorithms, the model simultaneously predicts the probability of enterprise entry, annual output value after entry, and industrial chain synergy index. This fully utilizes shared information between tasks, improving prediction accuracy and enabling multi-dimensional quantitative evaluation of the matching relationship between enterprises and park facilities. An improved slime mold optimization algorithm is introduced to automatically optimize the weight vector of the investment recommendation evaluation function, avoiding the subjectivity and blindness of manually setting weights. This algorithm combines chaotic initialization, slime mold optimization mechanisms, and gray wolf collaboration strategies, effectively balancing global search and local development capabilities. It can quickly and accurately find the optimal weight combination, thereby improving the quality of recommendation results. For multi-source heterogeneous data, BERT models and Min-Max normalization methods are used to process text and numerical data respectively, effectively extracting semantic and numerical features and solving the modeling challenges caused by data heterogeneity.

[0008] In one optional implementation, multi-source heterogeneous data from enterprises and park facilities within the smart park are collected, and feature extraction and embedding are performed to construct an enterprise feature matrix and a park resource feature matrix for the smart park, including: Collect multi-source heterogeneous data from enterprises and park facilities in the smart park. The multi-source heterogeneous data includes enterprise-side data of all enterprises and park-side data of all park facilities. For all text-based data in multi-source heterogeneous data, a pre-trained BERT model is used to extract the corresponding semantic feature vectors; For all numerical data in multi-source heterogeneous data, the Min-Max standardization method is used to normalize the numerical features and obtain the corresponding numerical feature vectors. The semantic feature vectors and / or numerical feature vectors of all enterprises in the enterprise-side data are concatenated to generate the corresponding enterprise feature matrix; The semantic feature vectors and / or numerical feature vectors of all park carriers in the park-side data are concatenated to generate the corresponding park resource feature matrix.

[0009] In one optional implementation, the park-side data includes registered capital, years of establishment, business scope, number of intellectual property rights, and administrative penalty records; The park-side data includes the building area, rental unit price, floor load-bearing capacity, and description of surrounding facilities; The registered capital, the number of years since establishment, the number of intellectual property rights, the area of ​​the building, the rental unit price, and the floor load-bearing capacity are all numerical data. The business scope, administrative penalty records, and surrounding facilities descriptions are all text-based data.

[0010] In one optional implementation, based on the enterprise feature matrix and park resource feature matrix of the smart park, a multi-task prediction model is constructed using machine learning algorithms to predict investment recommendation indicators for enterprises and park facilities, including: Based on the historical enterprise feature matrix and historical park resource feature matrix of the smart park, a multi-task prediction model is constructed with a model sample set containing real investment recommendation indicators as real labels. Each sample in the model sample set consists of the enterprise feature vector of any enterprise in the historical enterprise feature matrix and the park resource feature vector of any park carrier in the historical park resource feature matrix. We use machine learning algorithms to build an initial multi-task prediction model and define the loss function of the initial multi-task prediction model. Based on the model sample set, the initial multi-task prediction model is trained until the total loss value generated by the loss function is less than the total loss value threshold, thus obtaining the final multi-task prediction model, which is used to predict the investment recommendation indicators for enterprises and park carriers.

[0011] In one optional implementation, the investment promotion recommendation indicators include the probability of enterprises settling in, the annual output value of enterprises after settling in, and the industrial chain synergy index.

[0012] In one alternative implementation, the multi-task prediction model includes an input layer and at least three task prediction branches, each of which includes an expert network, a gating network, and a prediction head. The input layer is used to receive any enterprise feature vector from the input enterprise feature matrix and the corresponding park resource feature vector from the park resource feature matrix. The task prediction branch is used to predict the corresponding investment promotion recommendation indicators based on the enterprise feature vector and the park resource feature vector. An expert network is used to extract shared features from enterprise feature vectors and park resource feature vectors; Gated networks are used to perform weighted summation of the shared features output by expert networks to obtain weighted shared features; The prediction head is used to predict the corresponding investment recommendation indicators based on weighted shared features.

[0013] In one optional implementation, an investment recommendation evaluation function is constructed based on investment recommendation indicators, and an improved slime mold optimization algorithm is introduced to optimize the weight vector of the investment recommendation evaluation function, resulting in an optimized investment recommendation evaluation function, including: Using investment recommendation indicators as input, investment recommendation evaluation scores as output, and weight vectors as optimization objects, we construct an investment recommendation evaluation function. The model sample set with real investment recommendation indicators as real labels is used as the simulation environment, and the weight vector is used as the optimization object of the improved slime mold optimization algorithm, and encoded as the position vector of the individual in the improved slime mold optimization algorithm. Construct a fitness function for an improved slime mold optimization algorithm based on a simulated environment; Based on the fitness function, an improved slime mold optimization algorithm is used to iteratively optimize the object and obtain the optimal weight vector. Based on the optimal weight vector, the weight vector of the investment promotion recommendation evaluation function is adjusted to obtain the optimized investment promotion recommendation evaluation function.

[0014] In one alternative implementation, based on the fitness function, an improved slime mold optimization algorithm is used to iteratively optimize the object to obtain the optimal weight vector, including: The chaotic sequence is generated using a Logistic mapping, and then mapped to the solution space of individuals in the improved slime mold optimization algorithm to obtain the initial population. Using the fitness function, calculate the fitness value of each initial individual in the initial population, and select the top three individuals, the second best, and the third best individuals in terms of fitness from the initial population. Based on the standard position update mechanism of the slime mold optimization algorithm, the position of each initial individual in the initial population is updated to obtain the updated first position of the updated individual. Based on the gray wolf cooperative strategy, the position of each initial individual in the initial population is updated to obtain the updated second position of the updated individual. By introducing a random number selection, the first or second updated position is chosen as the final position of the corresponding updated individual, thus obtaining the updated population. Using the fitness function, calculate the fitness value of each updated individual in the updated population, and update the updated individual with the best fitness value as the best individual; The position update of the population is repeated. When the current iteration reaches the maximum number of iterations or the fitness value of the best individual meets the requirements, the iterative optimization of the population is terminated, and the position vector of the best individual at the current iteration is output. The optimal weight vector is obtained by decoding the position vector of the optimal individual.

[0015] In one optional implementation, based on the enterprise feature matrix and park resource feature matrix of the smart park, a multi-task prediction model and an optimized investment promotion recommendation evaluation function are used to generate investment promotion recommendation results for enterprises in the smart park, including: For each target enterprise in the smart park, extract the current enterprise feature vector from the current enterprise feature matrix of the smart park collected in real time. Based on the target company's current enterprise feature vector, we iterate through each current park resource feature vector in the real-time collected current park resource feature matrix, and use a multi-task prediction model to generate several current predictive investment recommendation indicators. Iterate through all current predicted investment attraction recommendation indicators, use the optimized investment attraction recommendation evaluation function to generate the target enterprise's investment attraction recommendation evaluation score for all park carriers; Based on the investment promotion recommendation evaluation scores, all park facilities are sorted in descending order of power, and a hard constraint rule is introduced for screening. Combined with the basic information of each park facility and the current predicted investment promotion recommendation indicators, an investment promotion recommendation list is obtained. By iterating through all enterprises in the smart park, and repeatedly performing multi-task prediction, investment recommendation evaluation, and sorting and filtering steps, an investment recommendation result is obtained that includes an investment recommendation list of all enterprises in the smart park.

[0016] Secondly, embodiments of the present invention provide a smart park investment promotion recommendation system based on machine learning, used to implement a smart park investment promotion recommendation method, the system comprising: The data acquisition unit is used to collect multi-source heterogeneous data from enterprises and park facilities in the smart park, and to extract and embed features to construct the enterprise feature matrix and park resource feature matrix of the smart park. The machine learning unit is used to construct a multi-task prediction model based on the enterprise feature matrix and park resource feature matrix of the smart park, using machine learning algorithms to predict investment recommendation indicators for enterprises and park facilities. The weight optimization unit is used to construct an investment recommendation evaluation function based on investment recommendation indicators, and introduce an improved slime mold optimization algorithm to optimize the weight vector of the investment recommendation evaluation function to obtain the optimized investment recommendation evaluation function. The investment promotion recommendation unit is used to generate investment promotion recommendation results for enterprises in the smart park based on the enterprise feature matrix and park resource feature matrix of the smart park, using a multi-task prediction model and an optimized investment promotion recommendation evaluation function.

[0017] A third aspect of this invention provides an electronic device, which includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, such that the at least one processor can perform the method proposed in the first aspect of the present invention.

[0018] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the steps of a smart park investment recommendation method based on machine learning, as provided in an embodiment of the present invention. Figure 3 This is a functional unit diagram of a smart park investment recommendation system based on machine learning provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0021] The present invention will be further described below with reference to the accompanying drawings.

[0022] Reference Figure 1 , Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention.

[0023] like Figure 1As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0024] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0025] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and an electronic program for a smart park investment recommendation system based on machine learning.

[0026] exist Figure 1 In the electronic device shown, the network interface 1004 is mainly used to conduct smart park investment promotion recommendations based on machine learning with the network server; the user interface 1003 is mainly used to interact with users; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device, and the electronic device calls the electronic program of the smart park investment promotion recommendation system based on machine learning stored in the memory 1005 through the processor 1001, and executes the smart park investment promotion recommendation method based on machine learning provided in the embodiment of the present invention.

[0027] Reference Figure 2 The present invention provides a smart park investment recommendation method based on machine learning, the method comprising: S201: Collect multi-source heterogeneous data of enterprises and park carriers in the smart park, and perform feature extraction and embedding to construct the enterprise feature matrix and park resource feature matrix of the smart park; S202: Based on the enterprise feature matrix and park resource feature matrix of the smart park, a multi-task prediction model is constructed using machine learning algorithms to predict the investment recommendation indicators for enterprises and park facilities. S203: Based on the investment recommendation indicators, construct an investment recommendation evaluation function, and introduce an improved slime mold optimization algorithm to optimize the weight vector of the investment recommendation evaluation function, thereby obtaining the optimized investment recommendation evaluation function; S204: Based on the enterprise feature matrix and park resource feature matrix of the smart park, use a multi-task prediction model and an optimized investment promotion recommendation evaluation function to generate investment promotion recommendation results for enterprises in the smart park.

[0028] The technical solution provided in this application has at least the following beneficial effects: By constructing a multi-task prediction model based on machine learning algorithms, the model simultaneously predicts the probability of enterprise entry, annual output value after entry, and industrial chain synergy index. This fully utilizes shared information between tasks, improving prediction accuracy and enabling multi-dimensional quantitative evaluation of the matching relationship between enterprises and park facilities. An improved slime mold optimization algorithm is introduced to automatically optimize the weight vector of the investment recommendation evaluation function, avoiding the subjectivity and blindness of manually setting weights. This algorithm combines chaotic initialization, slime mold optimization mechanisms, and gray wolf collaboration strategies, effectively balancing global search and local development capabilities. It can quickly and accurately find the optimal weight combination, thereby improving the quality of recommendation results. For multi-source heterogeneous data, BERT models and Min-Max normalization methods are used to process text and numerical data respectively, effectively extracting semantic and numerical features and solving the modeling challenges caused by data heterogeneity.

[0029] In one optional implementation, multi-source heterogeneous data from enterprises and park facilities within the smart park are collected, and feature extraction and embedding are performed to construct an enterprise feature matrix and a park resource feature matrix for the smart park, including: S2011: Collect multi-source heterogeneous data of enterprises and park carriers in the smart park from sources such as park management system, industrial and commercial database, and geographic information system (GIS) through API interface or direct database connection. The multi-source heterogeneous data includes enterprise-side data of all enterprises and park-side data of all park carriers. In this embodiment, the focus is on collecting the enterprise's business registration information (registered capital, years of establishment), operating status information (annual report revenue, tax payment, number of intellectual property rights), compliance information (administrative penalty records), and business scope information (business scope text). For missing values, numerical data is filled with the average value of similar enterprises, and text data is marked as "unknown". The key data collection includes physical space attributes (building area, floor height, load-bearing capacity), cost attributes (rental price per unit, property management fee), and geographical location attributes (description of surrounding amenities, such as distance to subway station, number of restaurants nearby, etc.). S2012: For all textual data in multi-source heterogeneous data, use a pre-trained model based on Bidirectional Encoder Representations from Transformers (BERT) to extract the corresponding semantic feature vectors. In this embodiment, textual data is input into the BERT model, and the [CLS] vector of the last hidden state is extracted as the semantic feature vector of the sentence. In order to reduce the dimensionality, a fully connected layer can be added to reduce the high-dimensional semantic vector (such as 768 dimensions) to a preset dimension (such as 64 dimensions) to obtain a dense semantic feature vector. S2013: For all numerical data in multi-source heterogeneous data, the Min-Max standardization method is used to normalize the numerical features and obtain the corresponding numerical feature vectors. In this embodiment, outliers in numerical data are detected, and data that deviates from more than 3 times the standard deviation are removed or corrected using the box plot method. The Min-Max standardization method is used to map all values ​​to the [0, 1] interval to eliminate the influence of units and obtain numerical feature vectors. S2014: Concatenate the semantic feature vectors and / or numerical feature vectors of all enterprises in the enterprise-side data to generate the corresponding enterprise feature matrix; S2015: Concatenate the semantic feature vectors and / or numerical feature vectors of all park carriers in the park-side data to generate the corresponding park resource feature matrix.

[0030] In one optional implementation, the park-side data includes registered capital, years of establishment, business scope, number of intellectual property rights, and administrative penalty records; The park-side data includes the building area, rental unit price, floor load-bearing capacity, and description of surrounding facilities; The registered capital, the number of years since establishment, the number of intellectual property rights, the area of ​​the building, the rental unit price, and the floor load-bearing capacity are all numerical data. The business scope, administrative penalty records, and surrounding facilities descriptions are all text-based data.

[0031] In one optional implementation, based on the enterprise feature matrix and park resource feature matrix of the smart park, a multi-task prediction model is constructed using machine learning algorithms to predict investment recommendation indicators for enterprises and park facilities, including: S2021: Based on the historical enterprise feature matrix and historical park resource feature matrix of the smart park, a multi-task prediction model is constructed with a model sample set containing real investment promotion recommendation indicators as real labels. Each sample in the model sample set consists of the enterprise feature vector of any enterprise in the historical enterprise feature matrix and the park resource feature vector of any park carrier in the historical park resource feature matrix. In this embodiment, information on the enterprises that have settled in and the carriers in which they have settled in is extracted from the historical investment promotion database as positive samples. The labels include: real settlement label (1), actual annual output value, and actual industrial chain synergy index. Using negative sampling technology, several non-resident carriers are randomly matched as negative samples for each resident enterprise. The labels are: real resident label (0), annual output value is recorded as 0 (or predicted value), and synergy index is recorded as calculated value. S2022: Using machine learning algorithms, construct an initial multi-task prediction model and define the loss function for the initial multi-task prediction model, as shown in the formula: In the formula, This represents the total loss value. These are the loss weighting coefficients; The loss value for the matching task is the error between the enterprise entry probability predicted in the model's investment promotion recommendation index and the actual entry label. The annual output value prediction loss is the error between the predicted annual output value after the enterprise moves in, as indicated by the model's investment promotion recommendation indicators, and the actual annual output value. This is the loss value of the industrial chain synergy, which is the error between the predicted industrial chain synergy index in the investment promotion recommendation indicators of the model and the actual synergy. S2023: Based on the model sample set, train the initial multi-task prediction model until the total loss value generated by the loss function is less than the total loss value threshold, and obtain the final multi-task prediction model, which is used to predict the investment recommendation indicators for enterprises and park carriers. In this embodiment, the Adam optimizer is used to iteratively update the model parameters until the total loss value on the validation set converges or the preset number of iterations is reached, and then the optimal model parameters are saved.

[0032] In one optional implementation, the investment promotion recommendation indicators include the probability of enterprises settling in, the annual output value of enterprises after settling in, and the industrial chain synergy index.

[0033] In one optional implementation, the multi-task prediction model adopts a multi-gate mixture-of-experts (MMoE) architecture, including an input layer and at least three task prediction branches, each of which includes an expert network, a gating network, and a prediction head. The input layer is used to receive any enterprise feature vector from the input enterprise feature matrix and the corresponding park resource feature vector from the park resource feature matrix. The task prediction branch is used to predict the corresponding investment promotion recommendation indicators based on the enterprise feature vector and the park resource feature vector. Several expert networks are set up, each of which is a multilayer perceptron (MLP) to extract shared features between enterprise feature vectors and park resource feature vectors; For each task (entry probability, annual output value, synergy index), a gating network is set up. The gating network dynamically calculates the weight coefficients of each expert network based on the input and outputs the weighted shared features. For example, the entry probability task may focus more on the business scope matching degree, while the annual output value task may focus more on the registered capital. The prediction head outputs predicted values ​​through a specific MLP layer, which are used to predict the corresponding investment recommendation indicators based on weighted shared features.

[0034] In one optional implementation, an investment recommendation evaluation function is constructed based on investment recommendation indicators, and an improved slime mold optimization algorithm is introduced to optimize the weight vector of the investment recommendation evaluation function, resulting in an optimized investment recommendation evaluation function, including: S2031: Using investment recommendation indicators as input, investment recommendation evaluation scores as output, and weight vectors as optimization objects, construct an investment recommendation evaluation function, the formula of which is: In the formula, For any selected target company With the j The evaluation score for investment promotion recommendations of the park's facilities; These are the weight coefficients in the weight vector; For any selected target company With the j The probability of companies moving into the industrial park; For any selected target company With the j Annual output value of enterprises after they move into the industrial park; For any selected target company With the j The industrial chain synergy index of the park's facilities; j Index for park infrastructure; The target company symbol can be arbitrarily selected; S2032: The model sample set with real investment recommendation indicators as real labels is used as the simulation environment, and the weight vector is used as the optimization object of the improved slime mold optimization algorithm, and encoded as the position vector of the individual in the improved slime mold optimization algorithm. The fitness function of the improved slime mold optimization algorithm based on the simulated environment is constructed as follows: In the formula, Weight vector X The corresponding fitness function; For any selected target company With the j Authentic investment attraction recommendation indicators for park facilities; N This represents the total number of facilities in the park. j Index for park infrastructure; The target company symbol can be arbitrarily selected; This is the cost penalty coefficient; For any selected target company With the j Normalized investment attraction costs for park facilities; This is the normalization function; Weight vector X The weighting coefficients in the text; S2034: Based on the fitness function, an improved slime mold optimization algorithm is used to iteratively optimize the object and obtain the optimal weight vector. ; S2035: Based on the optimal weight vector, adjust the weight vector of the investment promotion recommendation evaluation function to obtain the optimized investment promotion recommendation evaluation function, the formula of which is: In the formula, The optimal weight coefficients are those in the optimal weight vector. For any selected target company after optimization With the j The evaluation score for investment promotion recommendations of the park's facilities.

[0035] In one alternative implementation, an improved slime mold optimization algorithm is used based on the fitness function to iteratively optimize the object and obtain the optimal weight vector. ,include: S20341: Use Logistic mapping to generate chaotic sequences, and map the chaotic sequences to the solution space of individuals in the improved slime mold optimization algorithm to obtain the initial population; The formula is: In the formula, For the firstn+ 1. n There are several chaotic variables whose values ​​range from [0, 1]. The stability coefficient is typically 4. This sequence is ergodic and random, ensuring that the initial population is uniformly distributed in the solution space, avoiding getting trapped in local optima, which is superior to traditional random initialization. n Index for chaotic variables; In the formula, For the initial population, the first i An initial individual; For the first i One chaotic variable; These are the upper and lower bounds of the search space; i For individual indexes; S20342: Using the fitness function, calculate the fitness value of each initial individual in the initial population, and select the top three individuals, the second best individuals, and the third best individuals in the initial population based on their fitness ranking. S20343: A standard position update mechanism based on the slime mold optimization algorithm updates the position of each initial individual in the initial population to obtain the updated first position of the individual. The formula is as follows: In the formula, For the first t+ In the first iteration of the population, the first i The first position of the updated individual; For the first t In the population of the second iteration i Each updated individual is the same as the initial individual in the first iteration; The parameter is linearly decreasing, decreasing linearly from 1 to 0; t This represents the current iteration number; S20344: Based on the gray wolf cooperative strategy, the position of each initial individual in the initial population is updated to obtain the updated second position of the individual, using the following formula: In the formula, For the first t+ In the first iteration of the population, the first i The second position of the updated individual; For the first t The iteration of the ... i Each updated individual, in the initial iteration, For the initial population, the first i An initial individual; These are random weighting coefficients, and their sum is 1. For the first t+ 1, t The first, second, and third potential movement vectors of the next iteration; For the first t The next iteration Distance vectors between the best individual, the second-best individual, and the third-best individual; A random number between [0, 1]; For control coefficients and oscillation coefficients; For the first t The convergence factor of the next iteration; t This represents the current iteration number; For the first t The best, second-best, and third-best individuals in the next iteration; In the formula, These are the maximum and minimum values ​​of the convergence factor; This is the threshold for the number of iterations; S20345: Introduce the selection of random numbers, and select the updated first position or the updated second position as the final position of the corresponding updated individual to obtain the updated population. The formula is: In the formula, For the first t+ In the first iteration of the population, the first i The final location of the updated individual; To select a random number; To select the threshold parameter; To select weighting coefficients; In the formula, A random number between [0, 1]; For the first t The optimal individual in the next iteration fitness value; For the first t The iteration of the ... i A newer individual fitness value; For the first t The worst individual in the next iteration fitness value; S20346: Using the fitness function, calculate the fitness value of each updated individual in the updated population, and update the updated individual with the best fitness value as the best individual; S20347: Repeatedly update the position of the population. When the current iteration reaches the maximum number of iterations or the fitness value of the best individual meets the requirements, terminate the iterative optimization of the population and output the position vector of the best individual at the current iteration. S20348: Decode the position vector of the optimal individual to obtain the optimal weight vector. .

[0036] In one optional implementation, based on the enterprise feature matrix and park resource feature matrix of the smart park, a multi-task prediction model and an optimized investment promotion recommendation evaluation function are used to generate investment promotion recommendation results for enterprises in the smart park, including: S2041: For each target enterprise in the smart park, extract the current enterprise feature vector from the current enterprise feature matrix of the smart park collected in real time. S2042: Based on the target company's current enterprise feature vector, traverse each current park resource feature vector in the real-time collected current park resource feature matrix, and use a multi-task prediction model to generate several current predictive investment recommendation indicators. S2043: Iterate through all current predicted investment attraction recommendation indicators, use the optimized investment attraction recommendation evaluation function to generate the target company's investment attraction recommendation evaluation score for all park facilities. , For the current target company after optimization With the The evaluation score for investment promotion recommendations of the park's facilities; S2044: Based on the investment promotion recommendation evaluation scores, all park facilities are sorted in descending order of power, and a hard constraint rule is introduced for screening. Combined with the basic information of each park facility and the current predicted investment promotion recommendation indicators, an investment promotion recommendation list is obtained. In this embodiment, hard constraint rules include, for example: Area constraint: Enterprise's required area ≤ available space of the carrier; Environmental constraints: Enterprises with pollution levels ∈ the list of enterprises permitted to enter the park; Budget constraint: Enterprise's rental budget ≥ base rental price of the property; S2045: Traverse all enterprises in the smart park, repeatedly execute the multi-task prediction, investment recommendation evaluation, and sorting and filtering steps to obtain the investment recommendation results, which include an investment recommendation list of all enterprises in the smart park; In this embodiment, only the top-K park facilities in the investment promotion recommendation list are selected as the investment promotion recommendation results; The generated recommendation results are not only a list of carriers, but also include an interpretability analysis of the reasons for the recommendations; For example, for the top-ranked entity, the output would be: "Recommended score: 90 points. Key advantages: 85% predicted occupancy rate, expected annual output value contribution ranking in the top 10% of the park, and high degree of industrial chain synergy (strong matching with upstream and downstream enterprise A in the park)." Finally, the list of recommended investment projects is pushed to the terminal devices of the park's investment promotion management personnel to assist them in conducting precise investment promotion negotiations.

[0037] This invention also provides a smart park investment recommendation system based on machine learning, referring to... Figure 3 The device may include the following units: The data acquisition unit 301 is used to collect multi-source heterogeneous data of enterprises and park carriers in the smart park, and to extract and embed features to construct the enterprise feature matrix and park resource feature matrix of the smart park. Machine learning unit 302 is used to construct a multi-task prediction model based on the enterprise feature matrix and park resource feature matrix of the smart park using machine learning algorithms, and is used to predict the investment recommendation indicators for enterprises and park carriers. The weight optimization unit 303 is used to construct an investment recommendation evaluation function based on the investment recommendation indicators, and to introduce an improved slime mold optimization algorithm to optimize the weight vector of the investment recommendation evaluation function, thereby obtaining the optimized investment recommendation evaluation function. The investment promotion recommendation unit 304 is used to generate investment promotion recommendation results for enterprises in the smart park based on the enterprise feature matrix and park resource feature matrix of the smart park, using a multi-task prediction model and an optimized investment promotion recommendation evaluation function.

[0038] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. Memory, used to store computer programs; The processor, when executing the program stored in the memory, implements the machine learning-based smart park investment recommendation method of the present invention.

[0039] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned terminal and other devices. The memory can include Random Access Memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor.

[0040] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0041] Furthermore, to achieve the above objectives, embodiments of the present invention also propose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the machine learning-based smart park investment recommendation method of the embodiments of the present invention.

[0042] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable hardware devices (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0043] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0044] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0045] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0046] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. "And / or" indicates that either one or both can be chosen. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0047] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A smart industrial park investment recommendation method based on machine learning, characterized in that, The method includes: Collect multi-source heterogeneous data of enterprises and park facilities in the smart park, extract and embed features, and construct enterprise feature matrix and park resource feature matrix of the smart park. Based on the enterprise feature matrix and park resource feature matrix of the smart park, a multi-task prediction model is constructed using machine learning algorithms to predict investment recommendation indicators for enterprises and park facilities. Based on the investment recommendation indicators, an investment recommendation evaluation function is constructed, and an improved slime mold optimization algorithm is introduced to optimize the weight vector of the investment recommendation evaluation function, thus obtaining the optimized investment recommendation evaluation function. Based on the enterprise characteristic matrix and park resource characteristic matrix of the smart park, a multi-task prediction model and an optimized investment promotion recommendation evaluation function are used to generate investment promotion recommendation results for enterprises in the smart park.

2. The smart park investment recommendation method based on machine learning according to claim 1, characterized in that, Collect multi-source heterogeneous data from enterprises and park facilities in the smart park, extract and embed features to construct an enterprise feature matrix and a park resource feature matrix for the smart park, including: Collect multi-source heterogeneous data from enterprises and park facilities in the smart park. The multi-source heterogeneous data includes enterprise-side data of all enterprises and park-side data of all park facilities. For all text-based data in multi-source heterogeneous data, a pre-trained BERT model is used to extract the corresponding semantic feature vectors; For all numerical data in multi-source heterogeneous data, the Min-Max standardization method is used to normalize the numerical features and obtain the corresponding numerical feature vectors. The semantic feature vectors and / or numerical feature vectors of all enterprises in the enterprise-side data are concatenated to generate the corresponding enterprise feature matrix; The semantic feature vectors and / or numerical feature vectors of all park carriers in the park-side data are concatenated to generate the corresponding park resource feature matrix.

3. The smart park investment recommendation method based on machine learning according to claim 2, characterized in that, The data from the park includes registered capital, years of establishment, business scope, number of intellectual property rights, and administrative penalty records; The park-side data includes the building area, rental unit price, floor load-bearing capacity, and description of surrounding facilities; The registered capital, the number of years since establishment, the number of intellectual property rights, the area of ​​the building, the rental unit price, and the floor load-bearing capacity are all numerical data. The business scope, administrative penalty records, and surrounding facilities descriptions are all text-based data.

4. The smart park investment recommendation method based on machine learning according to claim 3, characterized in that, Based on the enterprise characteristic matrix and park resource characteristic matrix of the smart park, a multi-task prediction model is constructed using machine learning algorithms to predict investment recommendation indicators for enterprises and park facilities, including: Based on the historical enterprise feature matrix and historical park resource feature matrix of the smart park, a multi-task prediction model is constructed with a model sample set containing real investment recommendation indicators as real labels. Each sample in the model sample set consists of the enterprise feature vector of any enterprise in the historical enterprise feature matrix and the park resource feature vector of any park carrier in the historical park resource feature matrix. We use machine learning algorithms to build an initial multi-task prediction model and define the loss function of the initial multi-task prediction model. Based on the model sample set, the initial multi-task prediction model is trained until the total loss value generated by the loss function is less than the total loss value threshold, thus obtaining the final multi-task prediction model, which is used to predict the investment recommendation indicators for enterprises and park carriers.

5. The smart park investment recommendation method based on machine learning according to claim 4, characterized in that, The investment promotion recommendation indicators include the probability of enterprises settling in, the annual output value of enterprises after settling in, and the industrial chain synergy index.

6. The smart park investment recommendation method based on machine learning according to claim 5, characterized in that, The multi-task prediction model includes an input layer and at least three task prediction branches, each of which includes an expert network, a gating network, and a prediction head. The input layer is used to receive any enterprise feature vector from the input enterprise feature matrix and the corresponding park resource feature vector from the park resource feature matrix. The task prediction branch is used to predict the corresponding investment promotion recommendation indicators based on the enterprise feature vector and the park resource feature vector. An expert network is used to extract shared features from enterprise feature vectors and park resource feature vectors; Gated networks are used to perform weighted summation of the shared features output by expert networks to obtain weighted shared features; The prediction head is used to predict the corresponding investment recommendation indicators based on weighted shared features.

7. The smart park investment recommendation method based on machine learning according to claim 6, characterized in that, Based on the investment recommendation indicators, an investment recommendation evaluation function is constructed, and an improved slime mold optimization algorithm is introduced to optimize the weight vector of the investment recommendation evaluation function, resulting in an optimized investment recommendation evaluation function, including: Using investment recommendation indicators as input, investment recommendation evaluation scores as output, and weight vectors as optimization objects, we construct an investment recommendation evaluation function. The model sample set with real investment recommendation indicators as real labels is used as the simulation environment, and the weight vector is used as the optimization object of the improved slime mold optimization algorithm, and encoded as the position vector of the individual in the improved slime mold optimization algorithm. Construct a fitness function for an improved slime mold optimization algorithm based on a simulated environment; Based on the fitness function, an improved slime mold optimization algorithm is used to iteratively optimize the object and obtain the optimal weight vector. Based on the optimal weight vector, the weight vector of the investment promotion recommendation evaluation function is adjusted to obtain the optimized investment promotion recommendation evaluation function.

8. The smart park investment recommendation method based on machine learning according to claim 7, characterized in that, Based on the fitness function, an improved slime mold optimization algorithm is used to iteratively optimize the object and obtain the optimal weight vector, including: The chaotic sequence is generated using the Logistic mapping, and then mapped to the solution space of individuals in the improved slime mold optimization algorithm to obtain the initial population. Using the fitness function, calculate the fitness value of each initial individual in the initial population, and select the top three individuals, the second best, and the third best individuals in terms of fitness from the initial population. Based on the standard position update mechanism of the slime mold optimization algorithm, the position of each initial individual in the initial population is updated to obtain the updated first position of the updated individual. Based on the gray wolf cooperative strategy, the position of each initial individual in the initial population is updated to obtain the updated second position of the updated individual. By introducing a random number selection, the first or second updated position is chosen as the final position of the corresponding updated individual, thus obtaining the updated population. Using the fitness function, calculate the fitness value of each updated individual in the updated population, and update the updated individual with the best fitness value as the best individual; The position update of the population is repeated. When the current iteration reaches the maximum number of iterations or the fitness value of the best individual meets the requirements, the iterative optimization of the population is terminated, and the position vector of the best individual at the current iteration is output. The optimal weight vector is obtained by decoding the position vector of the optimal individual.

9. The smart park investment recommendation method based on machine learning according to claim 8, characterized in that, Based on the enterprise characteristic matrix and park resource characteristic matrix of the smart park, a multi-task prediction model and an optimized investment promotion recommendation evaluation function are used to generate investment promotion recommendation results for enterprises in the smart park, including: For each target enterprise in the smart park, extract the current enterprise feature vector from the current enterprise feature matrix of the smart park collected in real time. Based on the target company's current enterprise feature vector, we iterate through each current park resource feature vector in the real-time collected current park resource feature matrix, and use a multi-task prediction model to generate several current predictive investment recommendation indicators. Iterate through all current predicted investment attraction recommendation indicators, use the optimized investment attraction recommendation evaluation function to generate the target enterprise's investment attraction recommendation evaluation score for all park carriers; Based on the investment promotion recommendation evaluation scores, all park facilities are sorted in descending order of power, and a hard constraint rule is introduced for screening. Combined with the basic information of each park facility and the current predicted investment promotion recommendation indicators, an investment promotion recommendation list is obtained. By iterating through all enterprises in the smart park, and repeatedly performing multi-task prediction, investment recommendation evaluation, and sorting and filtering steps, an investment recommendation result is obtained that includes an investment recommendation list of all enterprises in the smart park.

10. A smart park investment promotion recommendation system based on machine learning, used to implement the smart park investment promotion recommendation method as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition unit is used to collect multi-source heterogeneous data from enterprises and park facilities in the smart park, and to extract and embed features to construct the enterprise feature matrix and park resource feature matrix of the smart park. The machine learning unit is used to construct a multi-task prediction model based on the enterprise feature matrix and park resource feature matrix of the smart park, using machine learning algorithms to predict investment recommendation indicators for enterprises and park facilities. The weight optimization unit is used to construct an investment recommendation evaluation function based on investment recommendation indicators, and introduce an improved slime mold optimization algorithm to optimize the weight vector of the investment recommendation evaluation function to obtain the optimized investment recommendation evaluation function. The investment promotion recommendation unit is used to generate investment promotion recommendation results for enterprises in the smart park based on the enterprise feature matrix and park resource feature matrix of the smart park, using a multi-task prediction model and an optimized investment promotion recommendation evaluation function.

Images