Artificial intelligence analysis method based on industry data

Abstract

The invention discloses an artificial intelligence analysis method based on industry data, and the method comprises the following steps: A, obtaining logistics data, carrying out the preprocessing of the logistics data, and building a logistics data topological graph; b, according to the weather prediction data and the in-transit parcel prediction number, using a neural network model to predict the transport capacity demand among the composite nodes in the logistics data topological graph; c, when the total logistics transport capacity occupancy in the logistics data topological graph of the prediction result is smaller than the rated logistics transport capacity, a path with the shortest time consumption is adopted to transport the parcels; otherwise, using the ant colony algorithm model to optimize and deploy the real-time transport capacity among the composite nodes in the logistics data topological graph and the real-time transport capacity in the composite nodes. The defects in the prior art can be overcome, and transport capacity distribution can be adjusted in real time according to the logistics state.

Description

Technical Field

[0001] The present invention relates to the technical field of logistics optimization configuration, and in particular to an artificial intelligence analysis method based on industry data. Background Art

[0002] With the development of the Internet and e-commerce, the logistics industry has experienced explosive growth. As a result, the fluctuation range of logistics volume in different periods has gradually increased. In order to fully utilize logistics capacity to digest the peak of logistics volume while reducing logistics costs as much as possible, a variety of logistics capacity analysis and adjustment methods have been proposed in the prior art. However, the existing methods are all based on historical data to allocate capacity for a period of time in the future, and cannot take into account the real-time working conditions of logistics status changes. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide an artificial intelligence analysis method based on industry data, which can solve the shortcomings of the existing technology and realize real-time adjustment of transportation capacity allocation according to the logistics status.

[0004] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.

[0005] An artificial intelligence analysis method based on industry data includes the following steps:

[0006] A. Obtain logistics data, pre-process the logistics data, and establish a logistics data topology map;

[0007] B. Based on weather forecast data and the predicted number of packages in transit, a neural network model is used to predict the transportation demand between the composite nodes in the logistics data topology graph;

[0008] C. When the total logistics capacity occupancy in the predicted logistics data topology is less than the rated logistics capacity, the shortest path is used for package transportation; otherwise, the ant colony algorithm model is used to optimize the real-time capacity between the composite nodes in the logistics data topology and the real-time capacity within the composite nodes.

[0009] Preferably, the logistics data includes logistics site distribution data, transit processing speed of logistics sites, logistics capacity of logistics routes between logistics sites, and real-time package quantity of logistics sites.

[0010] Preferably, preprocessing the logistics data includes the following steps:

[0011] The real-time package quantity change curve of the logistics site is fitted, and the logistics sites are clustered and grouped according to the linear correlation of the real-time package quantity change curve. Any logistics site in the same group has at least one logistics route connecting with other logistics sites in the same group, and the average logistics capacity of at least one logistics route in the same group has not reached saturation.

[0012] Preferably, the logistics data topology diagram includes several composite nodes, each composite node represents all logistics sites in the same group and the logistics capacity between the logistics sites, and the connecting lines between different composite nodes represent the average logistics capacity between the corresponding composite nodes.

[0013] Preferably, in step B, a neural network model is first established, and the neural network model includes an input layer, a convolution layer, a pooling layer, a denormalization layer, a fully connected layer and an output layer. A feedback loop is set between the fully connected layer and the convolution layer, an activation function unit is set on the feedback loop, and a weighting unit is set at the input end of the convolution layer; then the neural network model is trained using a training data set.

[0014] Preferably, when predicting the transportation capacity demand between each composite node in the logistics data topology diagram, the weather forecast data and the predicted number of packages in transit are input into the input layer. The input layer normalizes the data and adjusts the data structure to make it meet the data input requirements of the composite neural network model. The input layer sends the data to the convolution layer. In the weighted unit, the input data and the feedback data of the previous moment fed back through the feedback loop are weighted, and then input into the convolution layer together. The convolution layer extracts the feature matrix of the input data, and then performs dimensionality reduction through the pooling layer. The denormalization layer performs system parameter analysis on the feature matrix and outputs it to the fully connected layer. The fully connected layer optimizes the weights and biases of the system parameters through the back propagation algorithm. The optimized system parameters are output to the activation function unit and the output layer respectively. The activation function unit selects and outputs the system parameters. The output layer adjusts the data format and outputs the data to the target space.

[0015] Preferably, in step C, optimizing and allocating the real-time transport capacity between the composite nodes includes the following steps:

[0016] The connecting lines whose logistics capacity occupancy is less than the rated logistics capacity are extracted to form an optimal path set. The shipping composite node is used as the starting point of the ant colony, and the receiving composite node is used as the end point of the ant colony. The number of ant colonies is proportional to the square of the number of optimal paths. The initial value of the pheromone of the optimal path is inversely proportional to the occupancy of the corresponding optimal path. The pheromone volatility calculation formula is ρ=ρ0+ρ′, where ρ0 is the basic volatility and ρ′ is the weighted volatility. The size of the weighted volatility is inversely proportional to the average vehicle speed on the optimal path. The average vehicle speed is predicted by expert experience using traffic congestion status and weather forecast data. The heuristic information value of the optimal path is inversely proportional to the number of packages to be transferred at the composite node at the end point of the optimal path. The probability of the ant choosing a path is proportional to the product of the pheromone concentration and the heuristic information value. Each time the ant colony completes an optimization operation, the pheromone concentration and the heuristic information value on the optimal path are updated. When the maximum number of iterations is reached or the optimal solution is obtained, the iterative optimization is stopped, and the current optimal solution is used as the target parameter for real-time capacity allocation between composite nodes.

[0017] Preferably, in step C, optimizing and allocating the real-time transport capacity within the composite node includes the following steps:

[0018] When the logistics routes within a composite node have insufficient capacity, vehicles are deployed from logistics routes with unsaturated capacity to logistics routes with insufficient capacity; when the number of parcels to be transferred at a logistics site within a composite node exceeds a set threshold, parcels to be sent to the above-mentioned logistics site are forwarded to the nearest logistics site where the number of parcels to be transferred does not exceed the set threshold; when the capacity of all logistics routes within a composite node is saturated and / or the number of parcels to be transferred at all logistics sites within the composite node exceeds the set threshold, the composite node is temporarily deleted from the optimal path set, and when the capacity of all logistics routes within the composite node is not saturated and the number of parcels to be transferred at all logistics sites does not exceed the set threshold, the composite node is re-added to the optimal path set.

[0019] The beneficial effect brought about by adopting the above technical solution is that the present invention can realize rapid response and feedback control of logistics status, thereby improving the accuracy and timeliness of logistics capacity scheduling. DETAILED DESCRIPTION

[0020] Neural networks and ant colony algorithms are both relatively common artificial intelligence algorithm models, and are also used in logistics scheduling systems. However, neural networks require a large amount of training data and are not very robust when facing multi-variable logistics status prediction conditions. Ant colony algorithms cannot guarantee real-time performance when solving problems with multiple targets and rapidly changing states, and it is difficult to maintain diversity. Therefore, at this stage, the application of artificial intelligence algorithm models such as neural networks and ant colony algorithms in logistics scheduling systems is limited to the overall prediction of future logistics capacity demand. When it comes to real-time adjustment of logistics capacity, it is basically based on manual scheduling and retrieval of the closest expert experience records.

[0021] In order to give full play to the potential of the artificial intelligence algorithm model and achieve the purpose of adjusting the allocation of transportation capacity in real time according to the logistics status, the present invention first obtains logistics data (logistics data includes logistics site distribution data, logistics site transit processing speed, logistics capacity of logistics lines between logistics sites, and real-time package quantity of logistics sites), pre-processes the logistics data, and establishes a logistics data topology map. Then, based on weather forecast data and the predicted number of packages in transit, a neural network model is used to predict the transportation capacity demand between each composite node in the logistics data topology map. Finally, when the total logistics capacity occupancy in the logistics data topology map of the predicted result is less than the rated logistics capacity, the shortest path is used for package transportation; otherwise, an ant colony algorithm model is used to optimize the real-time transportation capacity between each composite node in the logistics data topology map and the real-time transportation capacity within the composite node.

[0022] Specifically, the preprocessing of logistics data includes the following steps:

[0023] The real-time package quantity change curve of the logistics site is fitted, and the logistics sites are clustered and grouped according to the linear correlation of the real-time package quantity change curve. Any logistics site in the same group has at least one logistics route connecting with other logistics sites in the same group, and the average logistics capacity of at least one logistics route in the same group has not reached saturation.

[0024] The logistics data topology diagram includes several composite nodes, each of which represents all logistics sites in the same group and the logistics capacity between the logistics sites. The connecting lines between different composite nodes represent the average logistics capacity between the corresponding composite nodes.

[0025] The topology structure we designed innovatively uses composite nodes to simplify the structure. A composite node is a set of logistics sites and their corresponding logistics routes. When optimizing logistics capacity, the overall deployment of the topology structure and the local deployment within the composite node can be achieved simultaneously, thereby effectively reducing the number and length of paths that the ant colony algorithm needs to optimize.

[0026] When making capacity forecasts, a neural network model is first established. The neural network model includes an input layer, a convolution layer, a pooling layer, a denormalization layer, a fully connected layer, and an output layer. A feedback loop is set between the fully connected layer and the convolution layer, an activation function unit is set on the feedback loop, and a weighted unit is set at the input end of the convolution layer. Then the neural network model is trained using the training data set. Then, when predicting the transportation capacity demand between each composite node in the logistics data topology diagram, the weather forecast data and the predicted number of packages in transit are input into the input layer. The input layer normalizes the data and adjusts the data structure to make it meet the data input requirements of the composite neural network model. The input layer sends the data to the convolution layer. In the weighted unit, the input data and the feedback data of the previous moment fed back through the feedback loop are weighted, and then input into the convolution layer together. The convolution layer extracts the feature matrix of the input data, and then performs dimensionality reduction through the pooling layer. The denormalization layer performs system parameter analysis on the feature matrix and outputs it to the fully connected layer. The fully connected layer optimizes the weights and biases of the system parameters through the back propagation algorithm. The optimized system parameters are output to the activation function unit and the output layer respectively. The activation function unit selects and outputs the system parameters. The output layer adjusts the data format and outputs the data to the target space.

[0027] In view of the characteristics of logistics capacity forecasting, we have specially optimized the structure of the neural network model. By setting up a feedback loop, the feedback data at the previous moment can have a lagged effect on the forecasting process at the next moment, thereby simulating the hysteresis of the logistics status and improving the prediction accuracy.

[0028] The optimization and allocation of real-time transport capacity between composite nodes includes the following steps:

[0029] The connecting lines whose logistics capacity occupancy is less than the rated logistics capacity are extracted to form an optimal path set. The shipping composite node is used as the starting point of the ant colony, and the receiving composite node is used as the end point of the ant colony. The number of ant colonies is proportional to the square of the number of optimal paths. The initial value of the pheromone of the optimal path is inversely proportional to the occupancy of the corresponding optimal path. The pheromone volatility calculation formula is ρ=ρ0+ρ′, where ρ0 is the basic volatility and ρ′ is the weighted volatility. The size of the weighted volatility is inversely proportional to the average vehicle speed on the optimal path. The average vehicle speed is predicted by expert experience using traffic congestion status and weather forecast data. The heuristic information value of the optimal path is inversely proportional to the number of packages to be transferred at the composite node at the end point of the optimal path. The probability of the ant choosing a path is proportional to the product of the pheromone concentration and the heuristic information value. Each time the ant colony completes an optimization operation, the pheromone concentration and the heuristic information value on the optimal path are updated. When the maximum number of iterations is reached or the optimal solution is obtained, the iterative optimization is stopped, and the current optimal solution is used as the target parameter for real-time capacity allocation between composite nodes. By changing the setting method of pheromone and heuristic information value in the ant colony algorithm, we make the ant colony fully consider the impact of different factors on path optimization during the optimization process, and facilitate rapid updates after each optimization is completed. It is precisely because we have solved the problem of slow update speed of the ant colony algorithm that the ant colony algorithm can not only be used in the prediction stage, but also can quickly adjust the next logistics capacity allocation according to the current real-time logistics capacity status, thus achieving a "two-pronged approach" of prediction and real-time adjustment.

[0030] The optimization and deployment of real-time transport capacity within the composite node includes the following steps:

[0031] When the logistics routes within a composite node have insufficient capacity, vehicles are deployed from logistics routes with unsaturated capacity to logistics routes with insufficient capacity; when the number of parcels to be transferred at a logistics site within a composite node exceeds a set threshold, parcels to be sent to the above-mentioned logistics site are forwarded to the nearest logistics site where the number of parcels to be transferred does not exceed the set threshold; when the capacity of all logistics routes within a composite node is saturated and / or the number of parcels to be transferred at all logistics sites within the composite node exceeds the set threshold, the composite node is temporarily deleted from the optimal path set, and when the capacity of all logistics routes within the composite node is not saturated and the number of parcels to be transferred at all logistics sites does not exceed the set threshold, the composite node is re-added to the optimal path set.

[0032] The above two optimization and allocation processes are carried out simultaneously and influence each other (that is, the results of the ant colony algorithm optimization affect the total output capacity and total input capacity of the composite node, and the real-time capacity optimization within the composite node affects the heuristic information value of the ant colony algorithm), thereby accelerating the two and improving the optimization speed.

[0033] The entire analysis and optimization method we designed is based on existing theories. By optimizing the topological structure of logistics capacity and improving the neural network model and ant colony algorithm, we have achieved rapid response to logistics status and rapid calculation of optimization and control parameters. This is a revolutionary breakthrough in the field of logistics capacity forecasting and optimization.

[0034] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An artificial intelligence analysis method based on industry data, characterized in that The following steps are involved: A. Obtain logistics data, pre-process the logistics data, and establish a logistics data topology map; B. Based on weather forecast data and the predicted number of packages in transit, a neural network model is used to predict the transportation demand between the composite nodes in the logistics data topology graph; C. When the total logistics capacity occupied in the logistics data topology diagram of the predicted result is less than the rated logistics capacity, the shortest path is used for package transportation; Otherwise, the ant colony algorithm model is used to optimize the deployment of real-time transport capacity between composite nodes and within composite nodes in the logistics data topology graph.

2. The artificial intelligence analysis method based on industry data according to claim 1 is characterized by: The logistics data includes logistics site distribution data, logistics site transit processing speed, logistics capacity of logistics routes between logistics sites, and real-time package quantity at logistics sites.

3. The artificial intelligence analysis method based on industry data according to claim 2 is characterized in that: The preprocessing of logistics data includes the following steps: The real-time package quantity change curve of the logistics site is fitted, and the logistics sites are clustered and grouped according to the linear correlation of the real-time package quantity change curve. Any logistics site in the same group has at least one logistics route connecting with other logistics sites in the same group, and the average logistics capacity of at least one logistics route in the same group has not reached saturation.

4. The artificial intelligence analysis method based on industry data according to claim 3 is characterized by: The logistics data topology diagram includes a plurality of composite nodes, each composite node represents all logistics sites in the same group and the logistics capacity between the logistics sites, and the connection lines between different composite nodes represent the average logistics capacity between the corresponding composite nodes.

5. The artificial intelligence analysis method based on industry data according to claim 4 is characterized in that: In step B, a neural network model is first established. The neural network model includes an input layer, a convolution layer, a pooling layer, a denormalization layer, a fully connected layer and an output layer. A feedback loop is set between the fully connected layer and the convolution layer, an activation function unit is set on the feedback loop, and a weighted unit is set at the input end of the convolution layer; then the neural network model is trained using the training data set.

6. The artificial intelligence analysis method based on industry data according to claim 5 is characterized by: When predicting the transportation capacity demand between each composite node in the logistics data topology diagram, the weather forecast data and the predicted number of packages in transit are input into the input layer. The input layer normalizes the data and adjusts the data structure to make it meet the data input requirements of the composite neural network model. The input layer sends the data to the convolution layer. In the weighted unit, the input data and the feedback data of the previous moment fed back through the feedback loop are weighted, and then input into the convolution layer together. The convolution layer extracts the feature matrix of the input data, and then performs dimensionality reduction through the pooling layer. The denormalization layer performs system parameter analysis on the feature matrix and outputs it to the fully connected layer. The fully connected layer optimizes the weights and biases of the system parameters through the back propagation algorithm. The optimized system parameters are output to the activation function unit and the output layer respectively. The activation function unit selects and outputs the system parameters. The output layer adjusts the data format and outputs the data to the target space.

7. The artificial intelligence analysis method based on industry data according to claim 6 is characterized in that: In step C, optimizing and allocating the real-time capacity occupancy between the composite nodes includes the following steps: The connecting lines whose logistics capacity occupancy is less than the rated logistics capacity are extracted to form an optimal path set. The shipping composite node is used as the starting point of the ant colony, and the receiving composite node is used as the end point of the ant colony. The number of ant colonies is proportional to the square of the number of optimal paths. The initial value of the pheromone of the optimal path is inversely proportional to the occupancy of the corresponding optimal path. The pheromone volatility calculation formula is ρ=ρ0+ρ′, where ρ0 is the basic volatility and ρ′ is the weighted volatility. The size of the weighted volatility is inversely proportional to the average vehicle speed on the optimal path. The average vehicle speed is predicted by expert experience using traffic congestion status and weather forecast data. The heuristic information value of the optimal path is inversely proportional to the number of packages to be transferred at the composite node at the end point of the optimal path. The probability of the ant choosing a path is proportional to the product of the pheromone concentration and the heuristic information value. Each time the ant colony completes an optimization operation, the pheromone concentration and the heuristic information value on the optimal path are updated. When the maximum number of iterations is reached or the optimal solution is obtained, the iterative optimization is stopped, and the current optimal solution is used as the target parameter for real-time capacity allocation between composite nodes.

8. The artificial intelligence analysis method based on industry data according to claim 7 is characterized in that: In step C, optimizing and allocating the real-time transport capacity within the composite node includes the following steps: When the logistics routes within a composite node have insufficient capacity, vehicles are deployed from logistics routes with unsaturated capacity to logistics routes with insufficient capacity; when the number of parcels to be transferred at a logistics site within a composite node exceeds a set threshold, parcels to be sent to the above-mentioned logistics site are forwarded to the nearest logistics site where the number of parcels to be transferred does not exceed the set threshold; when the capacity of all logistics routes within a composite node is saturated and / or the number of parcels to be transferred at all logistics sites within the composite node exceeds the set threshold, the composite node is temporarily deleted from the optimal path set, and when the capacity of all logistics routes within the composite node is not saturated and the number of parcels to be transferred at all logistics sites does not exceed the set threshold, the composite node is re-added to the optimal path set.