Knowledge graph-based algorithm reckoning path intelligent planning method and system

Abstract

The invention relates to the technical field of knowledge maps, and discloses an algorithm reckoning path intelligent planning method and system based on a knowledge map, and the method comprises a path node evaluation module which is used for obtaining historical features of reckoning nodes in the knowledge map, calculating an importance score, and dividing the reckoning nodes into two types according to the importance score and a set score threshold value; the reckoning nodes are entity nodes or relation nodes participating in algorithm reckoning path construction in a knowledge graph and comprise a starting point node, an intermediate node and an end point node, and the path reckoning probability prediction module is used for predicting the probability that reckoning is successfully completed by a reckoning path corresponding to a path request according to the path request of a current reckoning task and sending the reckoning probability to the current reckoning task. According to the method, importance scoring is performed on the calculation nodes through the path node evaluation module, high / low importance nodes are divided, the high importance nodes are preferentially guaranteed to participate in path planning, the reliability and efficiency of a calculation result are improved, and the stability of a knowledge graph calculation system is favorably maintained.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph technology, and in particular to an algorithm-based intelligent path planning method and system based on knowledge graphs. Background Technology

[0002] In knowledge graph applications, the quality of algorithmic path planning directly affects inference efficiency and result accuracy. By optimizing the inference path, key nodes can be quickly located in complex knowledge networks, reducing invalid path searches and improving system response speed.

[0003] Existing knowledge graph path planning methods typically employ fixed search rules, lacking the ability to dynamically assess the importance of inferred nodes. When dealing with large-scale knowledge graphs, it becomes difficult to identify high-value nodes and prioritize their paths, leading to low-importance nodes consuming excessive search resources and reducing inference efficiency. Furthermore, current technologies lack the concept of "minimum adjustments," often relying on global reordering during path optimization, which can easily disrupt the original node association logic and affect the interpretability of the inference results. In addition, current methods lack quantitative assessment of the resource matching degree of path requests and do not establish a dynamic adjustment mechanism for out-of-order paths. They fail to balance node order preservation with adjustment costs during path optimization, resulting in limited planning effectiveness and restricted system scalability.

[0004] Therefore, the aforementioned technical problems need to be solved. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, this invention proposes an algorithm-based intelligent path planning method and system based on knowledge graphs.

[0006] To solve the above-mentioned technical problems, the basic technical solution proposed by this invention is as follows:

[0007] A knowledge graph-based algorithm-based intelligent path planning system includes:

[0008] The path node evaluation module is used to obtain the historical features of the inferred nodes in the knowledge graph to calculate the importance score. Based on the set score threshold, the inferred nodes are divided into two categories according to the importance score.

[0009] The inference nodes are entity nodes or relation nodes in the knowledge graph that participate in the construction of the algorithm inference path, including starting nodes, intermediate nodes and ending nodes;

[0010] The path estimation probability prediction module is used to predict the probability that the estimation path corresponding to the current estimation task will be successfully completed based on the path request of the current estimation task, and obtain the estimation probability.

[0011] The dynamic path adjustment module is used to calculate the path search interval based on the estimated node importance score and estimated probability. When a path request is initiated, the path initiator delays the path search interval before sending the path request to the intelligent planning and optimization module.

[0012] The intelligent planning and optimization module receives path requests and knowledge graph node resource information, adjusts the inferred path to sort the path nodes, and plans the optimal inferred path. Specifically, it includes:

[0013] Set the threshold for out-of-order paths and the maximum number of adjustments;

[0014] Calculate the average number of out-of-order node pairs in the inferred path, and compare the average number with the out-of-order path threshold.

[0015] If the average number is less than the out-of-order path threshold, then output the planned and calculated path.

[0016] If the average number is greater than or equal to the misordered path threshold, the calculated path is adjusted according to the maximum number of adjustments to control the average number of misordered node pairs in the calculated path to be less than the misordered path threshold.

[0017] Preferably, the path node evaluation module is used to obtain the historical features of the inferred nodes in the knowledge graph to calculate their importance scores, and to divide the inferred nodes into two categories according to their importance scores based on a set scoring threshold, including:

[0018] The knowledge graph subgraph for this inference task consists of a path network composed of multiple related inference nodes; the path request includes the path start node, the end node, and the desired inference dimension.

[0019] The historical characteristics include the success rate of historical node estimation, the strength of node association, and the node resource utilization rate.

[0020] Weighting coefficients are set for the success rate of historical node estimation, the strength of node association, and the resource utilization rate of node, respectively.

[0021] For any given prediction node, calculate the weighted average of the node's historical prediction success rate, node association strength, and node resource utilization rate with their respective weight coefficients.

[0022] The weighted average result is recorded as the importance score of the calculated node;

[0023] Set a scoring threshold;

[0024] If the importance score is greater than or equal to the score threshold, the inferred node is a high importance node; if the importance score is less than the score threshold, the inferred node is a low importance node.

[0025] Preferably, the path estimation probability prediction module is used to predict the probability that the estimation path corresponding to the current estimation task will be successfully completed based on the path request of the current estimation task, including:

[0026] Obtain the node association priority of all candidate inference paths, sort the candidate inference paths from high to low according to the node association priority, and assign a sorting number to each candidate inference path;

[0027] Calculate the priority factor of the candidate inference path with sequence number m, specifically as follows:

[0028] Obtain the number M of all candidate inference paths;

[0029] calculate The result serves as the priority factor for the candidate inference path, where the priority factor equals 1 when m = 1, and equals 1 when m = M.

[0030] Preferably, the path estimation probability prediction module is used to predict the probability that the estimation path corresponding to the current estimation task will be successfully completed based on the path request of the current estimation task, and further includes:

[0031] Calculate the resource factors for the candidate inference path with sequence number m, specifically as follows:

[0032] Obtain the remaining node resources (R') in the knowledge graph when the candidate inference path is initiated. m );

[0033] The amount of node resources R required to obtain the candidate inference path m ;

[0034] calculate Where, when R' m ≥R m When the resource factor equals 1, and R' m <R m When the resource factor is less than 1;

[0035] Calculate the priority factor of the candidate inference path with sequence number m, multiply it by the resource factor, and use the result as the inference probability of the candidate inference path.

[0036] Preferably, the dynamic path adjustment module is used to calculate the path search interval based on the estimated node importance score and estimated probability, including:

[0037] Set minimum search interval T min and maximum search interval T max ;

[0038] For any path request:

[0039] Set high importance weights and low importance weights, where the high importance weight is greater than the low importance weight, and the sum of the high importance weight and the low importance weight is equal to 1;

[0040] If the core inference node included in the path request is a high-importance node, the product of the high-importance weight and the inference probability is calculated, and the result is used as the path priority P of the path request. p ;

[0041] If the core inference node included in the path request is a low-importance node, calculate the product of the low-importance weight and the inference probability, and use the result as the path priority P of the path request. p ;

[0042] The specific calculation steps for calculating the final path search interval for this path request are as follows: (T) max -P p ×(T max -T min ));

[0043] The moment when the path request is initiated is recorded as the initial moment;

[0044] The time interval for continuing the path search after calculating the initial time is recorded as the adjustment time. The control path initiator sends the path request to the intelligent planning and optimization module at the adjustment time.

[0045] A knowledge graph-based algorithm-based intelligent path planning system includes:

[0046] Node evaluation steps: Obtain the historical features of the inferred nodes in the knowledge graph, calculate the importance score, and divide the inferred nodes into two categories according to the set score threshold.

[0047] The inference nodes are entity nodes or relation nodes in the knowledge graph that participate in the construction of the algorithm inference path;

[0048] Probability prediction steps: Based on the path request of the current calculation task, predict the probability that the calculation path corresponding to the path request will be successfully completed, and obtain the calculation probability.

[0049] Dynamic adjustment steps: Calculate the path search interval based on the estimated node importance score and estimated probability. When a path request is initiated, the path initiator delays the path search interval before sending the path request to the planning and optimization stage.

[0050] The planning and optimization steps are as follows: Receive path requests and knowledge graph node resource information; adjust the inference path to sort the path nodes; and plan the optimal inference path. Specifically, this includes:

[0051] Set the threshold for out-of-order paths and the maximum number of adjustments;

[0052] Calculate the average number of out-of-order node pairs in the inferred path, and compare the average number with the out-of-order path threshold.

[0053] If the average number is less than the out-of-order path threshold, then output the planned and calculated path.

[0054] If the average number is greater than or equal to the misordered path threshold, the calculated path is adjusted according to the maximum number of adjustments to control the average number of misordered node pairs in the calculated path to be less than the misordered path threshold.

[0055] Preferably, the node evaluation step includes:

[0056] For the knowledge graph subgraph of this inference task, multiple related inference nodes participating in path construction are identified;

[0057] Obtain the historical characteristics of the inferred nodes, including the historical inference success rate of the nodes, the node association strength, and the node resource utilization rate;

[0058] Configure weight coefficients for node historical estimation success rate, node association strength, and node resource utilization rate, respectively;

[0059] Calculate the weighted average of the historical features and corresponding weight coefficients of any inferred node, and use it as the importance score of that node;

[0060] Based on the set scoring threshold, the inferred nodes are divided into high-importance nodes and low-importance nodes.

[0061] Preferably, the probability prediction step includes:

[0062] All candidate inference paths are sorted from high to low according to node association priority, and a sorting number is assigned to each candidate inference path;

[0063] Calculate the priority factor based on the order and total number of candidate inference paths;

[0064] Obtain the remaining node resources and required node resources of the candidate inference path, and calculate the resource factor;

[0065] The product of the priority factor and the resource factor is used as the probability of the candidate inference path.

[0066] Preferably, the dynamic adjustment step includes:

[0067] Set the minimum and maximum search intervals for pathfinding;

[0068] Based on the core inferred node type included in the path request, select high importance weight or low importance weight, and calculate path priority by combining inferred probability;

[0069] Calculate the path search interval based on path priority, minimum search interval, and maximum search interval;

[0070] Adjust the sending time of path requests according to the path search interval, and send the path requests to the planning and optimization stage at the adjusted time.

[0071] Preferably, the planning optimization step further includes:

[0072] Collect all path requests and number the corresponding estimated paths according to the order of receipt; record any estimated path as the target path and record the estimated path with a number less than the target path number as the preceding path;

[0073] Compare the path priorities of the target path and the preceding path. If the priority of the target path is greater than that of the preceding path, then record the two as an out-of-order node pair.

[0074] Count the number of out-of-order node pairs in all inferred paths and calculate the average number;

[0075] When the average number is greater than or equal to the threshold of the misordered path, the path with the largest number of misordered nodes is adjusted, including swapping the order of nodes in the path or replacing nodes.

[0076] Repeat the adjustment process until the average number is less than the out-of-order path threshold or the maximum number of adjustments is reached, and output the planned predicted path.

[0077] The beneficial effects of this invention are:

[0078] This invention uses a path node evaluation module to score the importance of inferred nodes, classifying them into high / low importance nodes. Prioritizing high-importance nodes in path planning improves the reliability and efficiency of the inference results, contributing to the stability of the knowledge graph inference system. In the probability prediction step, the inference probability is calculated by combining the node association priority and resource matching degree of candidate inferred paths, aligning with the core logic of "association strength priority, resource matching secondary" in knowledge graph inference, thus improving the prediction accuracy of successful path inference. A dynamic path adjustment module calculates the path search interval; paths corresponding to high-importance nodes have high priority and short search intervals, allowing them to enter the planning stage first. Low-importance nodes have longer path request intervals, reducing resource consumption and optimizing resource allocation in the time dimension, maintaining efficient system operation. By identifying and locally adjusting misordered node pairs, the problem of path order disorder caused by network latency or dynamic node changes is solved, preventing high-priority paths from being blocked by low-priority paths and reducing the time cost of ineffective searches. This invention sets a misordered path threshold and a maximum number of adjustments, controlling adjustment costs while allowing a certain degree of misorder, balancing path optimization effects with system resource consumption, and avoiding performance loss due to excessive adjustments. Attached Figure Description

[0079] Figure 1 This is a system block diagram of the knowledge graph-based algorithm-based intelligent path planning system of the present invention.

[0080] Figure 2 This is a flowchart of the knowledge graph-based algorithm for intelligent path planning according to the present invention. Detailed Implementation

[0081] The following will be combined with the appendix Figure 1 To be continued Figure 2 The technical solutions in the embodiments of the present invention have been clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0082] A knowledge graph-based intelligent path planning system for algorithm inference includes: a path node evaluation module, used to obtain historical features of inference nodes in the knowledge graph to calculate importance scores, and classify inference nodes into two categories according to their importance scores based on a set scoring threshold; the inference nodes are entity nodes or relation nodes in the knowledge graph that participate in the construction of algorithm inference paths, including start nodes, intermediate nodes, and end nodes; a path inference probability prediction module, used to predict the probability of successful completion of the inference path corresponding to the current inference task based on the path request, and obtain the inference probability; and a dynamic path adjustment module, used to calculate the path search interval based on the importance score and inference probability of the inference nodes, and adjust the path search interval when an inference node is selected. When a path request is initiated, the path initiator delays the path search interval before sending the path request to the intelligent planning and optimization module. The intelligent planning and optimization module receives the path request and knowledge graph node resource information, adjusts the inferred path to sort the path nodes, and plans the optimal inferred path. Specifically, this includes: setting a misordered path threshold and a maximum number of adjustments; calculating the average number of misordered node pairs in the inferred path and comparing the average number with the misordered path threshold; if the average number is less than the misordered path threshold, outputting the planned inferred path; if the average number is greater than or equal to the misordered path threshold, adjusting the inferred path according to the maximum number of adjustments to control the average number of misordered node pairs in the inferred path to be less than the misordered path threshold.

[0083] Optionally, the path node evaluation module is used to obtain the historical features of the inferred nodes in the knowledge graph to calculate their importance scores. Based on a set scoring threshold, the inferred nodes are divided into two categories according to their importance scores: a knowledge graph subgraph for the current inference task, consisting of a path network composed of multiple related inference nodes; the path request includes a path start node, an end node, and a desired inference dimension; the historical features include the node's historical inference success rate, node association strength, and node resource utilization rate; weight coefficients are set for the node's historical inference success rate, node association strength, and node resource utilization rate; for any inferred node, a weighted average of the node's historical inference success rate, node association strength, and node resource utilization rate with their respective weight coefficients is calculated; the weighted average result is recorded as the importance score of the inferred node; a scoring threshold is set; if the importance score is greater than or equal to the scoring threshold, the inferred node is a high-importance node; if the importance score is less than the scoring threshold, the inferred node is a low-importance node.

[0084] Optionally, the path estimation probability prediction module is used to predict the probability that the estimation path corresponding to the current estimation task will successfully complete the estimation based on the path request of the current estimation task. This includes: obtaining the node association priorities of all candidate estimation paths; sorting the candidate estimation paths according to their node association priorities from high to low, with each candidate estimation path corresponding to a sorting number; calculating the priority factor of the candidate estimation path with sorting number m, specifically: obtaining the number M of all candidate estimation paths; calculating... The result serves as the priority factor for the candidate inference path, where the priority factor equals 1 when m = 1, and equals 1 when m = M.

[0085] Optionally, the path estimation probability prediction module is used to predict the probability that the estimation path corresponding to the current estimation task will be successfully completed based on the path request of the current estimation task. It further includes: calculating the resource factor of the candidate estimation path with sequence number m, specifically: obtaining the remaining node resource amount R' in the knowledge graph when the candidate estimation path is initiated. m ; Obtain the node resource amount R required for this candidate inference path m ;calculate Where, when R' m ≥R m When the resource factor equals 1, and R' m <R m When the resource factor is less than 1, the priority factor of the candidate inference path with the sequence number m is calculated and multiplied by the resource factor, and the result is used as the inference probability of the candidate inference path.

[0086] Optionally, the dynamic path adjustment module is used to calculate the path search interval based on the estimated node importance score and estimated probability, including: setting a minimum search interval T. min and maximum search interval T max For any path request: assign a high importance weight and a low importance weight, where the high importance weight is greater than the low importance weight, and the sum of the high importance weight and the low importance weight equals 1; if the core inference node included in the path request is a high importance node, calculate the product of the high importance weight and the inference probability, and use the result as the path priority P of that path request. p If the core inference node included in the path request is a low-importance node, calculate the product of the low-importance weight and the inference probability, and use the result as the path priority P of the path request. p ; Calculate the final path search interval for this path request. The specific calculation steps are as follows: T max -P p ×(T max -T min The moment when the path request is initiated is recorded as the initial moment; the moment when the path search interval continues after the initial moment is recorded as the adjustment moment, and the path initiator sends the path request to the intelligent planning and optimization module at the adjustment moment.

[0087] A knowledge graph-based intelligent path planning method for algorithm inference includes: a node evaluation step: obtaining historical features of inference nodes in the knowledge graph, calculating importance scores, and classifying inference nodes into two categories according to their importance scores based on a set scoring threshold; the inference nodes are entity nodes or relation nodes in the knowledge graph that participate in the construction of algorithm inference paths; a probability prediction step: predicting the probability of successful completion of the inference path corresponding to the current inference task based on the path request, obtaining the inference probability; and a dynamic adjustment step: calculating the path search interval based on the importance score and inference probability of the inference nodes, and adjusting the path search interval when a path request is initiated. After delaying the path search interval, the path request is sent to the planning and optimization stage. The planning and optimization steps are as follows: receive the path request and knowledge graph node resource information, adjust the inferred path to sort the path nodes, and plan the optimal inferred path. Specifically, this includes: setting the misordered path threshold and the maximum number of adjustments; calculating the average number of misordered node pairs in the inferred path and comparing the average number with the misordered path threshold; if the average number is less than the misordered path threshold, output the planned inferred path; if the average number is greater than or equal to the misordered path threshold, adjust the inferred path according to the maximum number of adjustments to control the average number of misordered node pairs in the inferred path to be less than the misordered path threshold.

[0088] The node evaluation steps include: identifying multiple associated inference nodes participating in path construction for the knowledge graph subgraph of this inference task; obtaining the historical characteristics of the inference nodes, including the node's historical inference success rate, node association strength, and node resource utilization rate; configuring weight coefficients for the node's historical inference success rate, node association strength, and node resource utilization rate respectively; calculating the weighted average of the historical characteristics and corresponding weight coefficients of any inference node as the importance score of that node; and classifying the inference nodes into high-importance nodes and low-importance nodes based on the set scoring threshold.

[0089] The probability prediction step includes: sorting all candidate inference paths from high to low according to node association priority, and assigning a sorting number to each candidate inference path; calculating a priority factor based on the sorting number and total number of candidate inference paths; obtaining the remaining node resources and required node resources of the candidate inference paths, and calculating a resource factor; and using the product of the priority factor and the resource factor as the inference probability of the candidate inference path.

[0090] The dynamic adjustment steps include: setting the minimum and maximum search intervals for path search; selecting high or low importance weights based on the core inference node types included in the path request, and calculating the path priority in combination with the inference probability; calculating the path search interval based on the path priority, minimum search interval, and maximum search interval; adjusting the sending time of the path request according to the path search interval, and sending the path request to the planning and optimization stage at the adjusted time.

[0091] The planning optimization steps further include: collecting all path requests and numbering the predicted paths corresponding to the path requests in the order of receipt; designating any predicted path as the target path and the predicted path with a number less than the target path number as a preceding path; comparing the path priorities of the target path and the preceding path, and if the target path priority is greater than the preceding path priority, then the two are designated as a misordered node pair; counting the number of misordered node pairs for all predicted paths and calculating the average number; when the average number is greater than or equal to the misordered path threshold, selecting the predicted path with the largest number of misordered node pairs for adjustment, including swapping the order of nodes in the path or replacing nodes; repeating the adjustment process until the average number is less than the misordered path threshold or the maximum number of adjustments is reached, and outputting the planned predicted path.

[0092] The specific implementation method is as follows:

[0093] Example 1: A knowledge graph-based algorithm for intelligent path planning and deduction, comprising:

[0094] The working process of the path node evaluation module:

[0095] For the investment risk assessment and calculation task of a certain financial knowledge graph, the knowledge graph subgraph includes calculation nodes such as "industry nodes", "company nodes", and "financial indicator nodes".

[0096] Historical characteristics include: the success rate of historical node estimation (such as the accurate matching rate of the "Listed Company A" node in past risk assessments), the strength of node association (such as the business association weight between "Banking Industry" and "Listed Company A"), and the node resource utilization rate (such as the frequency of simultaneous requests to the node).

[0097] The weights are set as follows: historical prediction success rate weight 0.4, node association strength weight 0.4, and node resource utilization rate weight 0.2.

[0098] For the node “Listed Company A”, the historical prediction success rate is 0.92×0.4 + correlation strength is 0.85×0.4 + resource utilization rate is 0.7×0.2 = 0.868. The scoring threshold is set at 0.75. Therefore, “Listed Company A” is a high-importance node.

[0099] For the node “Listed Company B”, the historical prediction success rate is 0.65×0.4 + correlation strength is 0.5×0.4 + resource utilization rate is 0.8×0.2 = 0.58 < 0.75, therefore it is a low importance node.

[0100] The working process of the path estimation probability prediction module:

[0101] There are 6 candidate paths (M=6) for a certain investment risk assessment path request. After sorting them according to the node association priority, the candidate path with the sequence number m=3 is arranged as follows:

[0102] Priority factor = 1 - (3 - 1) / 6 = 1 - 2 / 6 ≈ 0.67;

[0103] The required node resources for this path are R3 = 15 units, the current remaining node resources are R'3 = 20 units, and the resource factor is min(1,20 / 15) = 1.

[0104] The estimated probability is 0.67 × 1 = 0.67.

[0105] The working process of the dynamic path adjustment module:

[0106] Set the minimum search interval Tmin = 0.4 seconds and the maximum search interval Tmax = 6 seconds;

[0107] High importance is weighted at 0.7, and low importance is weighted at 0.3;

[0108] The path request containing the highly important node "Listed Company A" has a calculated probability of 0.85 and a path priority Pp = 0.7 × 0.85 = 0.595.

[0109] Path search interval = 6 - 0.595 × (6 - 0.4) = 6 - 0.595 × 5.6 = 6 - 3.332 = 2.668 seconds;

[0110] The initial time of the path request is t0, and the adjustment time is t0+2.668 seconds. At this time, it is sent to the intelligent planning and optimization module.

[0111] The working process of the intelligent planning and optimization module:

[0112] Collect 6 path requests, numbered 1-6 in the order of receipt, with corresponding path priorities of 0.35, 0.25, 0.6, 0.5, 0.45, and 0.2 respectively;

[0113] Calculate the number of misordered node pairs: the path priority of number 3 is 0.6 > number 1 (0.35) and number 2 (0.25), and the number of misordered pairs is 2; the path priority of number 4 is 0.5 > number 1 (0.35) and number 2 (0.25), and the number of misordered pairs is 2; the path priority of number 5 is 0.45 > number 1 (0.35) and number 2 (0.25), and the number of misordered pairs is 2; the average number of misordered pairs is (2+2+2) / 3 = 2;

[0114] Set the out-of-order path threshold to 1.2, and the maximum number of adjustments to 4;

[0115] The average number of pairs is greater than or equal to 1.2, so adjustment is needed. Find the paths with the largest number of misordered pairs (numbers 3, 4, and 5), and prioritize adjusting path 3.

[0116] Traverse the preceding requests (numbers 1 and 2), swap the node order of number 3 with number 1, and recalculate the average number of misordered pairs = 1.6; swap the node order of number 3 with number 2, the average number = 1.4; select number 2 corresponding to the smallest single replacement value of 1.4 as the target request;

[0117] After replacement, the average quantity is 1.4 ≥ 1.2, and the number of swaps is 1 < 4. Continue to adjust number 4:

[0118] Swap the order of nodes numbered 4 and 1. The average number of nodes is 1.0 < 1.2, so stop adjusting.

[0119] Output the calculated path after planning to complete intelligent planning.

[0120] Based on the disclosure and teachings of the foregoing specification, those skilled in the art can make changes and modifications to the above embodiments. Therefore, the present invention is not limited to the specific embodiments disclosed and described above, and some modifications and changes to the present invention should also fall within the protection scope of the claims of the present invention. Furthermore, although some specific terms are used in this specification, these terms are only for convenience of explanation and do not constitute any limitation on the present invention.

Claims

1. An algorithm-based path calculation intelligent planning system based on a knowledge graph, characterized in that, Comprise: The path node evaluation module is used for obtaining the historical characteristic calculation importance score of the inference node in the knowledge graph, and dividing the inference node into two categories according to the set score threshold value; The inference node is an entity node or a relationship node participating in the algorithm inference path construction in the knowledge graph, including a starting node, an intermediate node and a terminal node; The path inference probability prediction module is used for predicting the probability of successfully completing the inference of the inference path corresponding to the path request according to the path request of the current inference task, and obtaining the inference probability; The dynamic path adjustment module is used for calculating the path search interval according to the importance score of the inference node and the inference probability, and sending the path request to the intelligent planning optimization module after delaying the path search interval when the path request is initiated; The intelligent planning optimization module is used for receiving the path request and the knowledge graph node resource information, adjusting the inference path to sort the path nodes, and planning the optimal inference path, specifically comprising: Setting the out-of-order path threshold value and the maximum adjustment number; Comparing the average number with the out-of-order path threshold value; If the average number is less than the out-of-order path threshold value, the planned inference path is output; If the average number is greater than or equal to the out-of-order path threshold value, the inference path is adjusted according to the maximum adjustment number to control the average number of out-of-order node pairs in the inference path to be less than the out-of-order path threshold value. 2.The knowledge graph-based algorithmic path-reasoning intelligent planning system of claim 1, wherein, The path node evaluation module is used for obtaining the historical characteristic calculation importance score of the inference node in the knowledge graph, and dividing the inference node into two categories according to the set score threshold value, comprising: The knowledge graph subgraph for this inference task is composed of a plurality of associated inference nodes to form a path network; the path request includes a path starting node, a terminal node and an expected inference dimension; The historical characteristics include node historical inference success rate, node association strength and node resource occupation rate; The weight coefficients of the node historical inference success rate, the node association strength and the node resource occupation rate are set respectively; For any one inference node, the weighted average of the node historical inference success rate, the node association strength and the node resource occupation rate and the respective weight coefficients is calculated; The weighted average result is recorded as the importance score of the inference node; The score threshold value is set; If the importance score is greater than or equal to the score threshold value, the inference node is a high importance node; if the importance score is less than the score threshold value, the inference node is a low importance node. 3.The knowledge graph based algorithmic path reasoning intelligent planning system according to claim 1, characterized in that, The path inference probability prediction module is used for predicting the probability of successfully completing the inference of the inference path corresponding to the path request according to the path request of the current inference task, comprising: Obtain the node association priority of all candidate inference paths, and sort the candidate inference paths according to the node association priority from high to low, each candidate inference path corresponding to a ranking number; The priority factor of the candidate inference path with the ranking number m is calculated, specifically: Obtain the number M of all candidate inference paths; Computing The result is a priority factor for the candidate estimated path, where when m = 1 the priority factor equals 1 and when m = M the priority factor equals 4.The knowledge graph-based algorithmic path-reasoning intelligent planning system of claim 3, wherein, The path inference probability prediction module is used for predicting the probability of successfully completing the inference of the inference path corresponding to the path request according to the path request of the current inference task, and further comprising: The resource factor of the candidate calculation path with the arrangement serial number m is calculated, and specifically, the resource factor of the candidate calculation path with the arrangement serial number m is calculated. Obtain the remaining node resources (R') in the knowledge graph when the candidate inference path is initiated. m ); an amount of node resources R required to obtain the candidate calculated path m ; Computing wherein, when R' m ≥ R m , the resource factor is equal to 1, and when R' m < R m , the resource factor is less than 1; The priority factor of the candidate calculation path with the arrangement serial number m is multiplied by the resource factor, and the result is taken as the calculation probability of the candidate calculation path. 5.The knowledge graph based algorithmic path reasoning intelligent planning system according to claim 1, wherein, The dynamic path adjustment module is used to calculate the path search interval according to the importance score of the calculation node and the calculation probability, and includes: Setting a minimum search interval T min and a maximum search interval T max ; For any one path request: Set a high importance weight and a low importance weight, wherein the high importance weight is greater than the low importance weight, and the sum of the high importance weight and the low importance weight is equal to 1; If the core node included in the path request is a high importance node, the product of the high importance weight and the estimated probability is calculated, and the result is taken as the path priority P of the path request p ; If the core node included in the path request is a low importance node, the product of the low importance weight and the estimated probability is calculated, and the result is taken as the path priority P of the path request p ; calculating the final path search interval of the path request, the specific calculation steps are as follows: (T max -P p × (T max -T min )); The time when the path request is initiated is recorded as the initial time; The time when the path search interval is continued after the initial time is recorded as the adjustment time, and the path initiation end is controlled to send the path request to the intelligent planning optimization module at the adjustment time.

6. A knowledge graph-based algorithm-based intelligent path planning method, applied to the knowledge graph-based algorithm-based intelligent path planning system as described in any one of claims 1-5, characterized in that, Including: The node evaluation step: obtaining the historical characteristics of the calculation node in the knowledge graph to calculate the importance score, and dividing the calculation node into two categories according to the set score threshold according to the importance score; The calculation node is an entity node or a relationship node in the knowledge graph participating in the algorithm calculation path construction; The probability prediction step: according to the path request of the current calculation task, the probability of successfully completing the calculation of the calculation path corresponding to the path request is predicted to obtain the calculation probability; The dynamic adjustment step: according to the importance score of the calculation node and the calculation probability, the path search interval is calculated, and when the path request is initiated, the path initiation end delays the path search interval and sends the path request to the planning optimization link; The planning optimization step: receiving the path request and the knowledge graph node resource information, adjusting the calculation path to sort the path nodes, planning the optimal calculation path, specifically including: Set the out-of-order path threshold and the maximum adjustment number; Calculate the average number of out-of-order node pairs in the calculation path, and compare the average number with the out-of-order path threshold; If the average number is less than the out-of-order path threshold, the planned calculation path is output; If the average number is greater than or equal to the out-of-order path threshold, the calculation path is adjusted according to the maximum adjustment number to control the average number of out-of-order node pairs in the calculation path to be less than the out-of-order path threshold. 7.The knowledge graph-based algorithmic path-reasoning intelligent planning system of claim 6, wherein, The node evaluation step includes: For the knowledge graph subgraph of the current calculation task, a plurality of associated calculation nodes participating in path construction are determined; Obtain the historical characteristics of the calculation node, the historical characteristics including node historical calculation success rate, node association strength and node resource occupation rate; Respectively configure weight coefficients for the node historical calculation success rate, the node association strength and the node resource occupation rate; Calculate the weighted average value of the historical characteristics of any calculation node and the corresponding weight coefficient as the importance score of the node; Based on the set score threshold, the calculation nodes are divided into high importance nodes and low importance nodes. 8.The knowledge graph based algorithmic path reasoning intelligent planning system of claim 1, wherein, The probability prediction step includes: Sort all candidate calculation paths according to node association priority from high to low, and assign each candidate calculation path an arrangement serial number; Based on the arrangement serial number and the total number of the candidate calculation paths, the priority factor is calculated; Obtain the remaining node resource quantity and the required node resource quantity of the candidate calculation path, and calculate the resource factor; The product of the priority factor and the resource factor is taken as the calculation probability of the candidate calculation path. 9.The knowledge graph based algorithmic path reasoning intelligent planning system according to claim 1, wherein, The dynamic adjustment step comprises: Setting a minimum search interval and a maximum search interval for path search; Based on the core node type contained in the path request, selecting a high importance weight or a low importance weight, and calculating the path priority in combination with the calculation probability; According to the path priority, the minimum search interval and the maximum search interval, the path search interval is calculated; According to the path search interval, the sending time of the path request is adjusted, and the path request is sent to the planning optimization link at the adjusted time. 10.The knowledge graph based algorithmic path reasoning intelligent planning system according to claim 1, wherein, The planning optimization step further comprises: Collecting all path requests, numbering the calculated paths corresponding to the path requests according to the receiving order, recording any calculated path as a target path, and recording the calculated paths with smaller numbers than the target path as front paths; Comparing the path priorities of the target path and the front paths, if the target path priority is greater than the front path priority, recording the two as a wrong sequence node pair; Counting the number of wrong sequence node pairs of all calculated paths, and calculating the average number; When the average number is greater than or equal to the wrong sequence path threshold, the calculated path with the largest number of wrong sequence node pairs is obtained for adjustment, which includes interchanging the node order in the path or replacing the node; Repeat the adjustment process until the average number is less than the wrong sequence path threshold or the maximum adjustment number is reached, and output the planned calculated path.

Images