Medicine development scheduling system capable of reducing cost and improving efficiency

Abstract

The invention relates to the technical field of medical research and development, in particular to a cost-reducing and efficiency-increasing drug development scheduling system, which comprises a data classification module, a data analysis module and a development scheduling optimization module, the time sequence features and the data features are deeply fused to construct multi-dimensional distribution features, development data of the same type are judged, a distributed storage technology is adopted for classification, and a hash value is allocated to each storage node; the data analysis module performs standardized processing on different types of development data, defines the development data as standard data, and judges whether the standard data have correlation or not; the development scheduling optimization module perceives the standard data with correlation to construct a standard tree diagram; and perceiving the development tree diagram, mapping and comparing the standard tree diagram, and if difference nodes exist in the development tree diagram and the standard tree diagram, calculating the influence degree.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical research and development, in particular, to a drug development scheduling system with cost reduction and efficiency improvement. BACKGROUND

[0002] At present, due to the strict time sequence dependence of each stage of drug development (such as target discovery, compound synthesis, clinical trials, etc.), the output results of the previous stage are the necessary prerequisite for starting the next stage. In order to avoid problems such as disorder of process sequence and disconnection of stage linkage in scheduling, the research and development personnel will usually build a development tree diagram of the drug to be developed before drug development, so as to intuitively present the causal logic time sequence dependence of the stage data such as "target discovery→compound synthesis→activity screening", and then the research and development personnel need to associate and analyze the historical data with the development tree diagram nodes, compare the execution sequence of each stage with the preset time sequence, check the causal relationship of the data time stamp, and check whether the indicators meet the starting standard; if the historical data is inconsistent with the logic of the tree diagram, the node level needs to be adjusted, the dependency relationship needs to be supplemented, or the parameter transmission rule needs to be corrected, so as to ensure that the tree diagram is consistent with the actual time sequence and provides a reliable basis for scheduling, avoiding logical errors affecting the development efficiency and success rate.

[0003] With the development of Internet technology, drug development data has exploded, providing a large amount of data from different sources (real-time experimental data, patient multi-omics data, historical project data, etc.) for development scheduling. A large amount of data from different sources not only enriches the decision basis, but also can mine potential rules through cross-validation, for example, comparing real-time experimental data with historical project data can quickly identify the activity trend of the compound; using patient multi-omics data can accurately locate the target and optimize the development path.

[0004] However, at the same time, due to the differences in format, dimension and time base of the massive data from different sources, the difficulty of data integration increases dramatically, and the research and development personnel need to spend a lot of time analyzing different source data and comparing the matching relationship between different source data and development tree diagram nodes, so that the research and development personnel need to spend a lot of time analyzing the data to analyze whether the development tree diagram is correct. In view of this, we propose a drug development scheduling system with cost reduction and efficiency improvement. SUMMARY

[0005] The purpose of the present application is to solve the problem that each stage of drug development has time sequence dependence, research and development personnel need to build a development tree diagram and associate historical data to verify the logic; Internet technology makes the development data explode, multi-source data (real-time experiment, patient multi-omics, historical project data, etc.) enrich the decision basis, which can be cross-validated to mine rules; but the differences in format, dimension and time base of the development data make the integration difficult, and it is necessary to spend time analyzing the development data and comparing the matching relationship with the tree diagram nodes to determine the correctness of the tree diagram, which affects the efficiency of development scheduling.

[0006] To achieve the above object, the present application provides a drug development scheduling system for cost reduction and efficiency improvement, comprising a data classification module, a data analysis module and a development scheduling optimization module, wherein:

[0007] The data classification module is used to construct time series features and data features, deeply fuse time series features and data features to construct multi-dimensional distribution features, and judge the same type of development data. Distributed storage technology is used for classification, and multiple types of development data are stored and classified. Each storage node is assigned a hash value;

[0008] The data analysis module uses a standardization processing method to process different types of development data, and defines the standardized development data as standard data, and judges whether the standard data has correlation. The development scheduling optimization module perceives the standard data with correlation, and uses time series method to construct standard tree diagram. The development tree diagram is perceived, and a hierarchical progressive mapping channel is constructed. The standard tree diagram is mapped and compared. If there is a difference node between the development tree diagram and the standard tree diagram during mapping and comparison, the influence degree of the difference node on the drug development process is calculated. The difference node is supplemented through supplementary logic, and the development data of the same node in the standard tree diagram is called out through the hash value in the data storage unit. The development data and the same node in the development tree diagram establish a mapping relationship.

[0009] As a further improvement of the technical solution, the data classification module comprises a data storage unit and a distributed storage execution unit. The data storage unit perceives development data from multiple sources and records the time, respectively constructs time series features and data features, deeply fuses time series features and data features to construct multi-dimensional distribution features, and judges the same type of development data in multiple sources of development data. Different types of development data are classified, and a hash function is used to assign a hash value to each storage node.

[0010] The distributed storage execution unit perceives multiple types of development data, and uses distributed storage technology to build a data warehouse to store multiple types of development data.

[0011] As a further improvement of the technical solution, the data storage unit constructs the multi-dimensional distribution feature by: perceiving the record time sequence of each source development data, calculating the interval time of adjacent times in the record time sequence, and constructing the interval time into a time sequence feature according to the record time sequence; perceiving the development data of each source, adding all data in each development data, and then dividing by the number of data to calculate the mean value of data in each development data; calculating the difference between each value and the mean value, adding the squares of the differences, and then dividing the sum by the number of data minus one to calculate the variance; and then constructing the mean value and the variance into a data feature according to a fixed order; assigning different weight coefficients to the time sequence feature and the data feature, multiplying the time sequence feature and the data feature by the corresponding weight coefficients, and then adding them to obtain the multi-dimensional distribution feature constructed by deep fusion of the time sequence feature and the data feature.

[0012] As a further improvement of the technical solution, the data storage unit perceives the multi-dimensional distribution features of the development data of each source, and calculates the similarity between each multi-dimensional distribution feature in turn; a similarity threshold is set, and if the similarity between two multi-dimensional distribution features is > the similarity threshold, the development data of the two different sources is judged to be the same type of data.

[0013] The beneficial effects of the above further solution are that the data storage unit can accurately capture the data timing rhythm and distribution rule, provide a unified standard for classification and integration of different types of data, solve the problem of difficulty in integrating multi-source data due to format differences, greatly reduce the time of researchers analyzing data matching relationships, make data classification more scientific and efficient, provide a reliable foundation for subsequent data correlation and verification in drug development scheduling, improve the logical accuracy and efficiency of the development process, avoid the impact of different types of data confusion on development progress and success rate, and help to mine the potential rules of different types of data to optimize the development path.

[0014] On the basis of the above technical solution, the present application can also be improved as follows.

[0015] As a further improvement of the technical solution, the distributed storage execution unit uses a distributed storage technology that: perceives a plurality of different types of development data, creates a storage node for each type of data, and marks the created storage node, and re-creates for the next type of development data; and assigns a corresponding hash value to each storage node through a hash function: h=H(K), where k is a fixed length hash value, H is a hash function, and K is the multi-dimensional distribution feature of each type of development data in the data classification module.

[0016] The beneficial effect of the above further scheme is that, through the role of the distributed storage technology in the correlation analysis unit, different types of development data are perceived, storage nodes are created for each type and are marked, dynamic adaptation to the multi-type and explosive growth of data in drug development is achieved, single storage node overload is avoided, data storage is orderly classified by type, hash value allocation realizes accurate positioning of data, data retrieval efficiency is greatly improved, the time of researchers to find and integrate data is reduced, stable support is provided for quick retrieval and correlation of data in subsequent development scheduling, efficient use of multi-source data is ensured, data value is mined to optimize the development process, the risk of development delay caused by data storage disorder is reduced, and drug development scheduling is promoted from the storage layer to reduce costs and increase efficiency;

[0017] Based on the above technical scheme, the application can be further improved as follows.

[0018] As a further improvement of the technical scheme, the data analysis module includes a data standardization unit and a correlation analysis unit; the data standardization unit uses a standardization processing method to map different types of development data to the same scale, and the correlation analysis unit is used to determine whether there is correlation between different types of data.

[0019] The standardization processing method in the data standardization unit is: calculating the mean of the development data: summing all development data of a certain type of development data, and then dividing by the number of development data; calculating the standard deviation of the development data: calculating the difference of the mean of all development data of a certain type of development data, then squaring the difference and adding it, and then dividing the sum by the number of development data, and finally taking the square root of the result; using the development data minus the mean, and then dividing by the standard deviation to get the standardized development data.

[0020] As a further improvement of the technical scheme, the correlation analysis unit sequentially calculates the correlation coefficient r between the selected standard data and other standard data, and the value range of the correlation coefficient r is [-1, 1];

[0021] Set a correlation threshold a, if the absolute value of the correlation coefficient r is greater than the correlation threshold a, then it is determined that there is correlation between the selected standard data and other standard data, and if the correlation coefficient r is negative, then the selected standard data and other standard data are negatively correlated, otherwise they are positively correlated;

[0022] After sequentially calculating the correlation coefficient r between the selected standard data and other standard data, the selected standard data is marked as marked data, and then other standard data except the marked data is randomly selected again, the correlation coefficient r between the selected standard data and other standard data except the marked data is calculated again, and the calculation of the correlation coefficient r between all standard data is completed.

[0023] The beneficial effect of the further scheme is that the data standardization unit standardizes the development data by calculating the mean value and the standard deviation, and unifies the data scale; the correlation analysis unit calculates the correlation coefficient between the standard data in sequence, determines the correlation and the positive and negative correlation, marks, and iteratively calculates the full analysis; the role is to provide standardized and correlation mining capabilities for drug development data, cooperate with the data classification module, adapt the classified data to subsequent analysis, support tree map mapping and difference processing unit tree map construction and difference analysis, accurately find data correlation, optimize scheduling logic, reduce errors caused by unclear data differences and correlations, improve development efficiency and success rate, and help reduce costs and increase efficiency.

[0024] Based on the above technical solutions, the application can also be improved as follows.

[0025] As a further improvement of the technical solution, the development scheduling optimization module includes a standard tree map construction unit and a tree map mapping and difference processing unit, the standard tree map construction unit analyzes the causal relationship between two different types of data using time series method and constructs a standard tree map; the tree map mapping and difference processing unit is used for the development tree map and the standard tree map, if there is a difference node between the development tree map and the standard tree map, the influence degree of the difference node on the drug development process is calculated; and the difference node is supplemented by supplementary logic; for the same node in the development tree map and the standard tree map, the development data of the same node in the standard tree map is called out, and the development data and the same node in the development tree map are mapped.

[0026] The beneficial effect of the further scheme is that the standard tree map construction unit analyzes the causal relationship between different types of data using time series method and constructs a standard tree map, and the tree map mapping and difference processing unit processes the difference between the development tree map and the standard tree map, calculates the influence degree of the difference node, supplements the node, and correlates the same node data; the role is to build a standard logic framework and optimize the development tree map scheduling logic; cooperate with the data analysis module, based on the standardized and correlated data of the data analysis module, the standard tree map construction unit accurately constructs the causal tree map, and the tree map mapping and difference processing unit deeply analyzes the difference; for the same node, reuse the historical data to reduce the development trial and error; from framework construction to difference optimization, ensure the development process time sequence compliance and logic self-consistency, accelerate the development process, reduce the cost caused by scheduling errors and data reuse difficulties, improve the success rate and efficiency of drug development, and promote the landing of cost reduction and efficiency increase

[0027] Based on the above technical solutions, the application can also be improved as follows.

[0028] As a further improvement of the technical solution, the time series method in the standard dendrogram construction unit analyzes the causal relationship between two standard data in the relevant data set: there is a correlation between the standard data A and the standard data B, the recording time of the standard data A and the standard data B is called, and if the recording time of the standard data A is greater than the recording time of the standard data B, it is judged that the standard data A is in the upper node of the standard data B in the standard dendrogram, thereby constructing the standard dendrogram.

[0029] As a further improvement of the technical solution, the dendrogram mapping and difference processing unit takes the root node of the development dendrogram and the standard dendrogram as the starting point, and performs one-to-one mapping comparison layer by layer and node by node; during the mapping comparison process, if there is a difference node between the development dendrogram and the standard dendrogram, the correlation coefficient r 父 , the correlation coefficient r 子 of the difference node with the last layer node and the next layer node in the data standardization unit is called, and then the correlation coefficient r 父 , the correlation coefficient r 子 is calculated to calculate the influence degree of the difference node on the drug development process: I = a · |r 子 | + β · |r 父 | + γ · L, wherein a and β are the correlation coefficient weights of the correlation coefficient r 子 and the correlation coefficient r 父 , γ is the level weight coefficient in the development dendrogram, L is the level in the development dendrogram, and I is the influence degree index.

[0030] The supplementary logic in the dendrogram mapping and difference processing unit: the last layer node and the next layer node of the difference node in the development dendrogram and the standard dendrogram are called, and if the last layer node and the next layer node are the same, the difference node in the development dendrogram is supplemented according to the difference node between the last layer node and the next layer node in the standard dendrogram.

[0031] The beneficial effects of the above further solution are that when there is a difference node between the development dendrogram and the standard dendrogram in the dendrogram mapping and difference processing unit, the upper and lower nodes of the difference node are called, if the upper and lower nodes are the same, the development dendrogram is supplemented according to the difference node between the upper and lower nodes in the standard dendrogram, the logical gap in the development dendrogram is accurately identified, the missing nodes in the development process are quickly supplemented with the mature time sequence logic of the standard dendrogram, and the time sequence causal relationship of the development dendrogram is complete and accurate, which provides a reliable logical basis for drug development scheduling, and avoids the disorder of scheduling order and the disconnection of stage connection caused by node loss.

[0032] In addition to the purposes, features and advantages described above, the present application has other purposes, features and advantages. The present application will be further described below with reference to the drawings. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 is the overall module schematic diagram of the present application;

[0034] Figure 2 is the working logic flow chart of the present application;

[0035] Figure 3 is the working principle flow chart of the standard tree diagram construction unit in the present application;

[0036] Figure 4 is the working principle flow chart of the tree diagram mapping and difference processing unit in the present application.

[0037] The meanings of various labels in the figure are as follows:

[0038] 100, data classification module; 110, data storage unit; 120, distributed storage execution unit; 200, data analysis module; 210, data standardization unit; 220, correlation analysis unit; 300, development scheduling optimization module; 310, standard tree diagram construction unit; 320, tree diagram mapping and difference processing unit. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0040] Reference Figures 1-4 As shown in the figure, a drug development scheduling system for cost reduction and efficiency improvement includes a data classification module 100, a data analysis module 200, and a development scheduling optimization module 300, wherein:

[0041] Because the process in the drug development process covers multiple stages such as target discovery, compound screening, and clinical trials, it has significant time sequence dependence (such as the next stage can only be started after the previous stage is completed) and data complexity (involving experimental data, resource allocation, compliance, and other multi-dimensional attributes), therefore the data classification module 100 includes a data storage unit 110 and a distributed storage execution unit 120;

[0042] The data storage unit 110 is used to classify drug development data from multiple sources, perceive development data from multiple sources, and record time, respectively corresponding to constructing time sequence features and data features, deeply fusing time sequence features and data features to construct multi-dimensional distribution features, and then judging the same type of development data in the development data from multiple sources and classifying them into one category;

[0043] The time sequence corresponding to the record of each source development data is perceived The interval time is The interval time is constructed into a time sequence feature in the existing order, that is The mean value of the data in each development data is calculated by perceiving the development data of each source: the mean value is obtained by adding all the data in each development data and dividing by the number of data

[0044] The difference between each value and the mean value is calculated, the sum of the squares of the differences is obtained, and the sum is divided by the number of data minus one, which is the variance Then the mean value and the variance are constructed into data features in a fixed order The mean value The central tendency of the data in each development data can be calculated, and the variance Var(x) can remind the dispersion degree of the data, so the time sequence feature and the data feature are deeply integrated to construct a multi-dimensional distribution feature: F = a · Y + b · X, where a and b are the weight coefficients corresponding to the time sequence feature Y and the data feature X, so that the multi-dimensional distribution feature F not only reveals the change rule of the development data in the time dimension, but also reveals its distribution characteristics in the attribute space, providing scientific and comprehensive feature support for intelligent scheduling, risk warning and decision optimization of the whole process of drug development.

[0045] The data storage unit 110 judges the development data of the same type in the multiple sources, and classifies it into a category, and the corresponding working principle is: perceiving the multi-dimensional distribution feature F of each source development data i , the similarity Distance(F i ,F j ) between each multi-dimensional distribution feature F i is calculated in turn: Where f ik and f jk are the multi-dimensional distribution features of two different source development data; set the similarity threshold value θ, if the similarity Distance(F i ,F j ) between the multi-dimensional distribution feature f ik and the multi-dimensional distribution feature f jk is greater than the similarity threshold value θ, then the development data of two different sources is judged as the same type data

[0046] The similarity Distance(F i ,F j) and compared with the similarity threshold θ, eliminating the subjectivity and ambiguity of manual judgment, and accurately and objectively identifying the same type of data with mathematical logic, making the classification of data from different sources more scientific and reliable, and improving the accuracy and efficiency of data-driven decision-making in the whole process of drug development, avoiding the confusion of development process or decision bias caused by misjudgment of data type.

[0047] In order to efficiently retrieve development data in the drug development scheduling system, the distributed storage execution unit 120 perceives multiple types of development data and uses distributed storage technology to build a data warehouse for storing multiple types of development data.

[0048] Distributed storage technology: perceives multiple different types of development data For each type of data t i Dynamically create storage nodes node i Form a node set For classifying and storing different development data to the corresponding storage node; and assigning a corresponding hash value to each storage node node i H = H(K), where k is a fixed length hash value, H is a hash function, and K is the multi-dimensional distribution feature of each type of development data in the data storage unit 110. The hash function can dynamically create storage nodes and assign hash values according to the number of development data, ensuring that the storage load of each node is relatively balanced; for example, when the amount of a certain type of data increases, the distributed storage execution unit 120 can automatically generate new nodes and assign corresponding hash values, avoiding the decline of storage performance caused by data overload in a single node, and achieving dynamic load balancing of distributed storage architecture; At the same time, the hash function generates a hash value based on the multi-dimensional distribution feature of the development data (combining time series features and data features), so that the same type of data (such as clinical trial data, compound screening data) is mapped to the same storage node, avoiding the mixed storage of different types of data, and improving the orderliness of the storage structure.

[0049] When analyzing multiple types of development data, multiple types of development data (such as size and material strength in medical device design parameters, raw material purity and reaction time in diagnostic reagent development) cannot be correlated between different types of development data due to dimensional (unit of measurement) and numerical distribution differences. Therefore, in order to eliminate the dimensional and numerical distribution differences of different types of development data, the data analysis module 200 includes a data standardization unit 210 and a correlation analysis unit 220. The data standardization unit 210 uses standardization processing method to map different types of development data to the same scale:

[0050] Calculate the mean and standard deviation of each development data: the mean of the development data is the sum of all development data in a certain type of development data, and then divided by the number of development data: Wherein μ is the mean value of a certain type of development data, is the sum of all development data of a certain type of development data, and n is the number of a certain type of development data;

[0051] The standard deviation of the development data is calculated by taking the difference between the mean value of all development data of a certain type of development data, squaring the difference, adding the squares, dividing the sum by the number of development data, and then taking the square root of the result. The standard deviation of a certain type of development data is:

[0052] Then, through the mean value and the standard deviation of a certain type of development data, each certain type of development data x i is linearly transformed to obtain the standardized development data x i : Wherein x i ' is the standardized development data, μ is the mean value of the development data before standardization, and σ is the standard deviation of the development data before standardization. Through the data standardization unit 210, the dimensional and numerical distribution difference barriers are broken down, and multiple types of data that cannot be directly correlated are analyzed under a unified standard to provide a basis for subsequent identification of potential relationships between data and optimization of development processes (such as adjusting reagent development parameters based on the correlation between material strength and reaction time).

[0053] The correlation analysis unit 220 perceives the standardized development data and defines it as standard data. A standard data is randomly selected, and the correlation coefficient r between the selected standard data and other standard data is calculated in turn: Wherein the value range of the correlation coefficient r is [-1, 1];

[0054] Set the correlation threshold a. If the absolute value of the correlation coefficient r is greater than the correlation threshold a, it is determined that there is a correlation between the selected standard data and other standard data. When the correlation coefficient r is negative, it is a negative correlation between the selected standard data and other standard data, and vice versa.

[0055] After calculating the correlation coefficient r between the selected standard data and other standard data, the selected standard data is marked as marked data. Then, a standard data other than the marked data is randomly selected again, and the correlation coefficient r between the selected standard data and other standard data is calculated again. This process is repeated until the correlation coefficient r between all standard data is calculated, thereby systematically mining the potential associations between multiple types of standard data. Whether it is different experimental indicators in drug development or various parameters in medical device development, the correlation coefficient can clearly present the correlation, such as determining the correlation between compound activity data and pharmacological indicators, providing a basis for finding key influencing factors and sorting out development logic (such as determining which data supports the start of "pharmacological and toxicological research").

[0056] The development scheduling optimization module 300 comprises a standard tree diagram construction unit 310 and a tree diagram mapping and difference processing unit 320. The standard tree diagram construction unit 310 is used for constructing standard data with perceived correlation into a standard tree diagram. Specifically, the cause-effect relationship between two standard data in a related data group is analyzed by time series method: it is perceived that there is correlation between standard data A and standard data B, the recording time of standard data A and standard data B is called, if the recording time of standard data A > the recording time of standard data B, it is judged that standard data A is in the upper node of standard data B in the standard tree diagram, forming a “parent node→child node” dependency relationship. By time series method, the data correlation is combed in objective time logic, a clear and accurate standard tree diagram is built for the drug development process, and the dependency relationship of each development stage is visualized. Data nodes such as target discovery and compound synthesis are arranged according to time causality. In subsequent scheduling, the rule of “preceding sequence completion before starting the following sequence” can be directly followed, avoiding repeated development or step misplacement due to logical confusion, ensuring reasonable development rhythm, and improving process standardization and controllability from the bottom logic.

[0057] The tree diagram mapping and difference processing unit 320 perceives the development tree diagram and constructs a hierarchical progressive mapping channel: taking the root nodes of the development tree diagram and the standard tree diagram as the starting point, one-to-one mapping comparison is performed layer by layer and node by node (such as root node→first layer child node→second layer child node…until leaf node); during the mapping comparison process, if there are difference nodes (nodes of the same level are different) in the development tree diagram and the standard tree diagram, the correlation coefficients r 父 and r 子 of the difference nodes in the data standardization unit 210 with the previous layer node and the next layer node respectively are called, and then the correlation coefficients r 父 and r 子 are used to determine the influence degree I=α·|r 子 |+β·|r 父 |+γ·L, where α and β are the correlation coefficient weights of the correlation coefficients r 子 and r 父 , γ is the hierarchical weight coefficient in the development tree diagram, L is the level in the development tree diagram, and I is the influence degree index.

[0058] By quantifying the correlation strength (absolute value of correlation coefficient) of the difference node with the upper and lower layer nodes and the level (influence range) into the influence degree I, the influence depth of the difference node on the subsequent compound synthesis and pharmacology and toxicology research stages can be determined through the influence degree I. If the influence degree I is large, resources are preferentially allocated to repair the difference, avoiding delay of the critical path; if the influence degree I is small, it can be handled in a non-critical period, so that the scheduling is consistent with the actual development logic, reducing invalid waiting and incorrect promotion, and reducing development cycle error.

[0059] And the tree diagram mapping and difference processing unit 320 supplements the difference node by supplementing logic: call the last layer node and the next layer node of the difference node in the development tree diagram and the standard tree diagram, if the last layer node and the next layer node are the same, then supplement the difference node in the development tree diagram according to the difference node between the last layer node and the next layer node in the standard tree diagram, avoid the subsequent scheduling confusion caused by the missing node, like misplacing "activity detection" in front of "structure modification", make the development strictly follow the scientific time sequence of "structure modification → activity detection → pharmacology and toxicology", and guarantee the development logic correct.

[0060] When the development tree diagram and the standard tree diagram have the same node, the hash value in the distributed storage execution unit 120 is used to call the development data of the same node in the standard tree diagram, and the mapping relationship between the development data and the same node in the development tree diagram is established, when the developer carries out drug development through the development tree diagram, the mapping relationship between the development data and the same node in the development tree diagram is used to provide intelligent assistance for the developer, when matching to the same node, the standard tree data is called through the hash value in the distributed storage execution unit 120 and the mapping is established, and the developer can directly reuse the high-quality data verified by history when scheduling.

[0061] The above shows and describes the basic principles, main features and advantages of the present application. It should be understood by those skilled in the art that the present application is not limited to the above-mentioned embodiments, the above-mentioned embodiments and descriptions in the specification are only preferred examples of the present application, and are not intended to limit the present application, various changes and improvements can be made without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A cost-effective drug development scheduling system, characterized by, The application relates to a drug development scheduling optimization system, which comprises a data classification module (100), a data analysis module (200) and a development scheduling optimization module (300), wherein: The data classification module (100) is used for constructing time sequence features and data features, deeply fusing the time sequence features and the data features to construct multi-dimensional distribution features, judging the same type of development data, and adopting a distributed storage technology to classify and store the development data of multiple types, and assigning a hash value to each storage node. The data analysis module (200) adopts a standardization processing method to process the development data of different types, defines the standardized development data as standard data, judges whether the standard data have correlation, the development scheduling optimization module (300) perceives the standard data having correlation, adopts a time sequence method to construct a standard tree diagram, perceives a development tree diagram, constructs a hierarchical progressive mapping channel, maps and compares the standard tree diagram, if there are difference nodes in the development tree diagram and the standard tree diagram during the mapping and comparison, calculates the influence degree of the difference nodes on the drug development process, supplements the difference nodes through a supplement logic, and calls the development data of the same nodes in the standard tree diagram through the hash value in the data storage unit (110), and establishes a mapping relationship between the development data and the same nodes in the development tree diagram.

2. The cost-effective drug development scheduling system of claim 1, wherein: The data classification module (100) comprises a data storage unit (110) and a distributed storage execution unit (120), the data storage unit (110) perceives development data of multiple sources and record times, respectively constructs time sequence features and data features, deeply fuses the time sequence features and the data features to construct multi-dimensional distribution features, judges the same type of development data in the development data of multiple sources, classifies the development data of different types, and assigns a hash value to each storage node through a hash function. The distributed storage execution unit (120) perceives development data of multiple types, and adopts a distributed storage technology to build a data warehouse to store the development data of multiple types.

3. A cost-effective drug development scheduling system according to claim 2, wherein: The data storage unit (110) constructs the multi-dimensional distribution features as follows: perceiving the record time sequence of the development data of each source, calculating the interval time of adjacent times in the record time sequence, constructing the interval time into time sequence features according to the record time sequence; perceiving the development data of each source, adding all data in each development data, and then dividing the sum by the number of data to calculate the mean value of data in each development data; calculating the difference value of each value and the mean value, adding the square of the difference value, dividing the sum by the number of data minus one to calculate the variance, and then constructing the mean value and the variance into data features according to a fixed sequence; assigning different weight coefficients to the time sequence features and the data features, multiplying the time sequence features, the data features and the corresponding weight coefficients, and then adding them, which are the multi-dimensional distribution features constructed by deeply fusing the time sequence features and the data features.

4. The cost-effective drug development scheduling system of claim 3, wherein: The data storage unit (110) perceives the multi-dimensional distribution characteristics of the development data from each source, and sequentially calculates the similarity between each multi-dimensional distribution characteristic; a similarity threshold is set, and if the similarity between two multi-dimensional distribution characteristics is > the similarity threshold, the development data from two different sources is judged as the same type of data.

5. The cost-effective drug development scheduling system of claim 3, wherein: The distributed storage execution unit (120) perceives a plurality of different types of development data, creates a storage node for each type of data, and marks the created storage node, and for the next type of development data, the storage node is recreated again; And through a hash function, each storage node is assigned a corresponding hash value: h = H (K), where k is a fixed length hash value, H is a hash function, and K is the multi-dimensional distribution characteristic of each type of development data in the data classification module (100).

6. The cost-effective drug development scheduling system of claim 1, wherein: The data analysis module (200) includes a data standardization unit (210) and a correlation analysis unit (220); the data standardization unit (210) uses a standardization processing method to map different types of development data to the same scale, and the correlation analysis unit (220) is used to determine whether there is a correlation between different types of data; The standardization processing method in the data standardization unit (210) is: calculating the mean of the development data: summing all development data in a certain type of development data, and then dividing by the number of development data; Calculate the standard deviation of the development data: calculate the difference of the mean of all development data in a certain type of development data, then square the difference and add it, divide the sum by the number of development data, and finally take the square root of the result; The development data is subtracted from the mean, and then divided by the standard deviation to obtain the standardized development data.

7. A cost-effective drug development scheduling system according to claim 6, wherein: The correlation analysis unit (220) sequentially calculates the correlation coefficient r between the selected standard data and other standard data, and the value range of the correlation coefficient r is [-1, 1]; Set a correlation threshold a, and if the absolute value of the correlation coefficient r is > the correlation threshold a, then it is determined that there is a correlation between the selected standard data and other standard data, and when the correlation coefficient r is negative, the selected standard data and other standard data are negatively correlated, otherwise they are positively correlated; After sequentially calculating the correlation coefficient r between the selected standard data and other standard data, the selected standard data is marked as marked data, and then other standard data excluding the marked data is randomly selected again, and the correlation coefficient r between the selected standard data and other standard data excluding the marked data is calculated again, until the correlation coefficient r between all standard data is calculated.

8. A cost-effective drug development scheduling system according to claim 7, wherein: The development scheduling optimization module (300) comprises a standard tree diagram construction unit (310) and a tree diagram mapping and difference processing unit (320), the standard tree diagram construction unit (310) adopts a time series method to analyze the causal relationship between two different types of data and constructs a standard tree diagram; the tree diagram mapping and difference processing unit (320) is used for, for the development tree diagram and the standard tree diagram, if there is a difference node between the development tree diagram and the standard tree diagram, the influence degree of the difference node on the drug development process is calculated; and the difference node is supplemented by supplementary logic; for the same node in the development tree diagram and the standard tree diagram, the development data of the same node in the standard tree diagram is called out, and the development data and the same node in the development tree diagram are mapped.

9. A cost-effective drug development scheduling system according to claim 8, wherein: The time series method in the standard tree diagram construction unit (310) analyzes the causal relationship between two standard data in a related data set: there is a correlation between the standard data A and the standard data B, the record time of the standard data A and the standard data B is called out, if the record time of the standard data A is greater than the record time of the standard data B, it is judged that the standard data A is in the upper node of the standard data B in the standard tree diagram, thereby constructing the standard tree diagram.

10. A cost-effective drug development scheduling system according to claim 9, wherein: The tree diagram mapping and difference processing unit (320) takes the root node of the development tree diagram and the standard tree diagram as the starting point, and compares and maps one by one layer by layer and node by node; In the mapping comparison process, if there is a difference node between the development tree and the standard tree, the correlation coefficient r 父 of the difference node with the node above and the node below in the data standardization unit (210) is called respectively 子 , and then the correlation coefficient r 父 , the correlation coefficient r 子 is calculated, and the influence degree of the difference node on the drug development process is calculated: I = α·|r 子 |+β·|r 父 |+γ·L, wherein α and β are the correlation coefficient weights of the correlation coefficient r 子 and the correlation coefficient r 父 , γ is the hierarchical weight coefficient in the development tree, L is the level in the development tree, and I is the influence degree index. The supplementary logic in the tree diagram mapping and difference processing unit (320): the last layer node and the next layer node of the difference node in the development tree diagram and the standard tree diagram are called out, if the last layer node and the next layer node are the same, the difference node in the development tree diagram is supplemented according to the difference node between the last layer node and the next layer node in the standard tree diagram.

Images