Investment management method and system based on life cycle

Abstract

The invention relates to the technical field of investment abnormal data detection, in particular to a life cycle-based investment management method and system, and the method comprises the steps: collecting various types of investment management data in a preset time period; determining the correlation degree between any type of investment management data and the rest of various types of investment management data; determining a constraint factor between the any type of investment management data and the rest any type of investment management data; obtaining a constraint correlation degree between any type of investment management data and the rest of various types of investment management data; obtaining various types of investment management data having strong relevance with any type of investment management data; evaluating the preference of each data point as a break point in any type of investment management data; and obtaining abnormal data in any type of investment management data based on the preference degree, removing the abnormal data, and performing investment management on various types of investment management data. Therefore, the security and reliability of investment decision making are improved.

Description

Technical Field

[0001] This application relates to the technical field of abnormal investment data detection, and particularly to an investment management method and system based on the life cycle. Background Art

[0002] Lifecycle Investment Management is a systematic method that takes the time dimension as the framework and dynamically adjusts investment strategies, risk preferences, and resource allocations according to the characteristics of the entire life cycle stage of the entity. Its core lies in regarding investment as a dynamic process rather than a static decision, emphasizing the time value, the matching of risk and return, and the optimization of the full-cycle cost and return. It can not only improve the financial health of individuals and institutions but also promote the transformation of resource utilization from "short-term consumption" to "long-term value creation".

[0003] When making project establishment decisions for investment projects in each stage of the life cycle, it is often necessary to integrate project research data and data of previous project materials in order to comprehensively evaluate investment projects and improve the scientificity of project investment through data analysis. However, in the process of collecting project research data and data of previous project materials, abnormal data usually inevitably exists, and directly analyzing the collected data will greatly affect the reliability of project decisions.

[0004] When using the Isolation Forest algorithm to detect abnormal data in project research data and data of previous project materials, the traditional algorithm randomly selects splitting features and splitting points, which greatly affects the construction efficiency and accuracy of isolation trees, and then affects the accuracy and efficiency of abnormal data detection, resulting in problems of investment decision deviation. Summary of the Invention

[0005] In order to solve the above technical problems, the purpose of this application is to provide an investment management method and system based on the life cycle, and the specific technical solutions adopted are as follows:

[0006] In the first aspect, an embodiment of this application provides an investment management method based on the life cycle, and the method includes the following steps:

[0007] Collect various types of investment management data within a preset time period;

[0008] Analyze the differences between adjacent data points in any type of investment management data, as well as the data distributions of the remaining types of investment management data within the acquisition time intervals corresponding to the adjacent data points, and determine the correlation degree between the any type of investment management data and the remaining types of investment management data;

[0009] Take the data at the same collection time in any one type of investment management data and any remaining type of investment management data as a feature point pair; based on the occurrence frequency of the feature point pair and the time difference between the feature point pairs, determine the constraint factor between any one type of investment management data and any remaining type of investment management data;

[0010] Combine the correlation degree and the constraint factor to obtain the constraint correlation degree between any one type of investment management data and the remaining various types of investment management data; classify the remaining various types of investment management data using the correlation between the constraint correlation degrees; obtain various types of investment management data that have a strong correlation with any one type of investment management data;

[0011] In the various types of investment management data with strong correlation, respectively obtain the relevant data points of each data point in any one type of investment management data. Take each data point and the relevant data point as split points, split the investment management data to which the split points belong, analyze the difference in the number of the two categories of data and the difference in the average level of the various types of investment management data after splitting, and combine the constraint correlation degree to evaluate the preference degree of each data point in any one type of investment management data as a split point;

[0012] Use the isolation forest algorithm to construct isolation trees for any one type of investment management data, select split points in the process of constructing the isolation trees based on the preference degree, obtain each abnormal data in any one type of investment management data and eliminate it, and perform investment management based on the various types of investment management data after eliminating the abnormal data.

[0013] In one embodiment, the determination of the correlation degree includes:

[0014] Calculate the ratio of the absolute value of the difference between adjacent data points in any one type of investment management data to the maximum value of the adjacent data points, denoted as the first ratio; count the number of data points in the remaining various types of investment management data that are not within the collection time interval of any one type of investment management data. Combine the first ratio, the number of data points, and the data distribution of the remaining various types of investment management data within the collection time interval corresponding to the adjacent data points in any one type of investment management data to calculate the correlation degree.

[0015] In one embodiment, the calculation method of the correlation degree is:

[0016] In the formula, GL QW is the correlation degree between Q - type investment management data and W - type investment management data, exp[] is the exponential function with the natural constant as the base, N Q is the number of data points of Q - type investment management data, The first ratio for the nth data point in the Q-type investment management data. Obtain all the data points of the W-type investment management data within the acquisition time interval corresponding to the nth data point and the (n + 1)th data point in the Q-type investment management data. Calculate the ratio of the extreme value to the maximum value of all the data points of the W-type investment management data, denoted as the second ratio. Calculate the degree of dispersion of all the data points of the W-type investment management data. is the product of the second ratio and the degree of dispersion. β is a preset value greater than 0, and n W is the number of data points in the W-type investment management data that are not within the acquisition time interval corresponding to the Q-type investment management data.

[0017] In one embodiment, the determination of the constraint factor includes:

[0018] Among all the feature point pairs of any type of investment management data and any remaining type of investment management data, calculate the ratio of the number of occurrences of each feature point pair to the total number of all feature point pairs, denoted as the third ratio. Calculate the cumulative sum of the time intervals between all occurrences of each feature point pair. The constraint factor is the fusion result of the third ratio and the cumulative sum.

[0019] In one embodiment, the constraint correlation is positively correlated with the correlation and negatively correlated with the constraint factor.

[0020] In one embodiment, the obtaining of various types of investment management data having a strong correlation with any type of investment management data includes:

[0021] Using the clustering algorithm to divide the constraint correlations between any type of investment management data and all the remaining types of investment management data into two categories, and taking the various types of investment management data in the category with the maximum average constraint correlation as the various types of investment management data having a strong correlation with any type of investment management data.

[0022] In one embodiment, the relevant data points are the data points in the various types of investment management data having a strong correlation that are closest to each data point in any type of investment management data in terms of acquisition time.

[0023] In one embodiment, the calculation method of the preference degree is:

[0024] where Y Qq is the preference degree for the data point q in the Q-type investment management data as a segmentation point; n q1 is the number of one type of data points after dividing the Q-type investment management data with the data point q as the segmentation point in the Q-type investment management data, and n q2 is the number of the other type of data points, is the mean value of one type of data points, is the mean value of another type of data points, R u is the constraint correlation degree between the Q - type investment management data and the u - type investment management data with strong correlation thereto; U is the number of types of investment management data with strong correlation to the Q - type investment management data, is the number of one type of data points after the u - type investment management data is segmented by the relevant data points of the data point q in the u - type investment management data, is the number of another type of data points, is the mean value of one type of data points, is the mean value of another type of data points.

[0025] In one embodiment, when the Q - type investment management data is segmented by using the data point q as the segmentation point, the data points in the Q - type investment management data that are greater than or equal to the data point q are taken as one type of data points, and the data points less than the data point q are taken as another type of data points; when the u - type investment management data is segmented by using the relevant data points of the data point q in the u - type investment management data, the same segmentation method as that when the Q - type investment management data is segmented by using the data point q as the segmentation point is adopted for segmentation;

[0026] The segmentation point for constructing the isolation tree based on the preference degree is: selecting the segmentation point with the maximum preference degree as the segmentation point for the isolation tree construction process.

[0027] In a second aspect, the embodiment of the present application further provides an investment management system based on the life cycle, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0028] The present application has at least the following beneficial effects:

[0029] This application collects various investment management data within a preset time period; analyzes the differences between adjacent data points in any type of investment management data, as well as the data distribution of the remaining types of investment management data within the collection time interval corresponding to the adjacent data points, to determine the correlation degree between the any type of investment management data and the remaining types of investment management data; the calculation of the correlation degree can improve the accuracy and robustness of anomaly detection when detecting anomalies in various investment management data, avoid the limitations of single-dimensional detection, and enhance the ability to capture complex dynamic patterns; uses the data at the same collection moment in the any type of investment management data and any remaining type of investment management data as a feature point pair; determines the constraint factor between the any type of investment management data and any remaining type of investment management data based on the occurrence frequency of the feature point pair and the time difference between the feature point pairs; combines the correlation degree and the constraint factor to obtain the constraint correlation degree between the any type of investment management data and the remaining types of investment management data; the determination of the constraint factor avoids the deviation in the calculation process of the correlation degree, introduces the frequency and time difference constraints of the feature point pair, eliminates false information that is statistically relevant but has no causal or business logic connection, and improves the accuracy of the correlation degree analysis; classifies the remaining types of investment management data using the correlation between the constraint correlation degrees; obtains various types of investment management data that have a strong correlation with the any type of investment management data; respectively obtains the relevant data points of each data point in the any type of investment management data from the various types of investment management data with strong correlation, uses the each data point and the relevant data point as split points respectively, splits the investment management data to which the split points belong, analyzes the number difference and the average level difference between the two categories of data after splitting, and combines the constraint correlation degree to evaluate the preference degree of each data point in the any type of investment management data as a split point; the preference degree reflects the suitability of each data point in various types of investment management data as a split point, avoids the problem of low anomaly data detection accuracy and efficiency caused by randomly selecting split points in the traditional isolation forest algorithm, and improves the accuracy of split point determination; constructs an isolation tree for the any type of investment management data using the isolation forest algorithm, selects the split points in the construction process of the isolation tree based on the preference degree, obtains each anomaly data in the any type of investment management data and eliminates it, and conducts investment management based on the various types of investment management data after eliminating the anomaly data, which improves the accuracy and efficiency of anomaly data detection in various types of investment management data, thereby improving the accuracy of investment management data and enhancing the reliability and security of investment decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] To more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0031] Figure 1 The flowchart of the steps of an investment management method based on the life cycle provided by an embodiment of the present application;

[0032] Figure 2 It is the flowchart of abnormal data detection. Detailed implementation manners

[0033] In order to further elaborate on the technical means and effects adopted by the present application to achieve the intended invention purpose, the following will, in conjunction with the accompanying drawings and preferred embodiments, elaborate in detail on an investment management method and system based on the life cycle proposed by the present application, its specific implementation manners, structures, features and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present application belongs.

[0035] The following will specifically describe the specific solutions of an investment management method and system based on the life cycle provided by the present application in conjunction with the accompanying drawings.

[0036] Please refer to Figure 1 , which shows the flowchart of the steps of an investment management method based on the life cycle provided by an embodiment of the present application. The method includes the following steps:

[0037] S1, collect various investment management data within a preset time period.

[0038] In the initial stage of the investment project in this embodiment, various investment research data of the investment project are obtained through market research and on-site inspections. Among them, the investment research data in this embodiment includes market scale data, growth rate data, investment cost data, and revenue prediction data. Secondly, obtain the actual execution data of the historical same-type projects of the investment project according to the previous project materials, that is, various historical material data, including project actual revenue data, cost overrun data, construction period delay data, risk loss data, and material resource data. All kinds of investment research data and various historical material data are recorded as various investment management data.

[0039] It should be noted that in this embodiment, the collection time period of various types of investment management data is one year. Taking a stage of the life cycle as an example for analysis, the types of investment research data and historical data can be set by the implementer according to the actual situation, and this embodiment does not limit it here.

[0040] Since different types of investment management data have different characteristics and formats, in this embodiment, all types of investment management data are numerically processed, that is, each type of investment management data is a numerical sequence. And because of the different types of investment management data, the collection frequencies of different types of investment management data may also be different, that is, the lengths of the numerical sequences of different types of investment management data are not exactly the same.

[0041] S2. Analyze the differences between adjacent data points in any type of investment management data, as well as the data distributions of the remaining types of investment management data within the collection time intervals corresponding to the adjacent data points, and determine the correlation degree between the any type of investment management data and the remaining types of investment management data.

[0042] When using the isolation forest algorithm to detect abnormal data in various types of investment management data, in the traditional isolation forest algorithm, the splitting points and splitting features are randomly selected when constructing the isolation tree, and the selection of the splitting points and splitting features directly affects the accuracy and efficiency of the isolation tree in detecting abnormal data points. Therefore, in this embodiment, the splitting points and splitting features are adaptively selected to improve the construction accuracy and efficiency of the isolation tree, and further improve the accuracy and efficiency of the isolation forest in detecting abnormal data.

[0043] When detecting abnormal data in various types of investment management data, this embodiment sets the number of isolation trees to 200, and each time 256 data points are sampled with replacement from the collected various types of investment management data as sample data. For any type of investment management data, here taking the Q type of investment management data as an example, for the Q type of investment management data, first obtain the associated data categories of the Q type of investment management data, and use the associated data categories as the features selected for each split to assist in the construction of the isolation tree, so that when each node of the isolation tree is split, the splitting accuracy and efficiency are maximized.

[0044] For all types of investment management data, obtain any type of investment management data other than the Q type of investment management data. Taking the W type of investment management data as an example, analyze the correlation between the Q type of investment management data and the W type of investment management data. The specific analysis process is as follows:

[0045] Calculate the absolute value of the difference between each data point in the Q - type investment management data and the subsequent adjacent data point, calculate the ratio of the absolute value of the difference to the maximum value of the two data points corresponding to the absolute value of the difference, and denote it as the first ratio; then, count the number of data points in the W - type investment management data that are not within the collection time zone corresponding to the Q - type investment management data. Combine the first ratio, the number of data points in the W - type investment management data that are not within the collection time zone corresponding to the Q - type investment management data, and the data distribution of the W - type investment management data within the collection time interval corresponding to adjacent data points in the Q - type investment management data to calculate the correlation degree between the Q - type investment management data and the W - type investment management data. The specific calculation method is as follows:

[0046] In the formula, GL QW is the correlation degree between the Q - type investment management data and the W - type investment management data, exp[] is the exponential function with the natural constant as the base, N Q is the number of data points in the Q - type investment management data, is the first ratio of the nth data point in the Q - type investment management data. Obtain all the data points of the W - type investment management data within the collection time interval corresponding to the nth data point and the (n + 1)th data point in the Q - type investment management data, calculate the ratio of the extreme value to the maximum value of all the data points of the W - type investment management data, and denote it as the second ratio. Calculate the degree of dispersion of all the data points of the W - type investment management data, is the product of the second ratio and the degree of dispersion, β is a preset value greater than 0, and n W is the number of data points in the W - type investment management data that are not within the collection time zone corresponding to the Q - type investment management data.

[0047] It should be noted that in this embodiment, β = 0.01, and its function is to avoid the denominator being 0. In this embodiment, the degree of dispersion is calculated using the standard deviation. Implementers can optionally choose other existing feasible calculation methods for the degree of dispersion, such as variance, coefficient of variation, etc.

[0048] It should be understood that the more the number of data points in the W - type investment management data that are not within the collection time zone corresponding to the Q - type investment management data, the less the overlapping part between the W - type investment management data and the collection time zone corresponding to the Q - type investment management data. That is, the more data points in the W - type investment management data are collected before or after the collection time zone of the Q - type investment management data, resulting in a smaller correlation between the W - type investment management data and the Q - type investment management data, that is, a smaller correlation degree between the Q - type investment management data and the W - type investment management data. reflects the numerical distribution of the nth data point and its adjacent data points in the Q - type investment management data. It reflects the numerical distribution of the nth data point in the Q-type investment management data corresponding to the data point in the W-type investment management data. If The greater the difference between and 1, the greater the difference in the numerical distribution of the data points at the same moment between the Q-type investment management data and the W-type investment management data, the smaller the correlation between the Q-type investment management data and the W-type investment management data, that is, the smaller the degree of association between the Q-type investment management data and the W-type investment management data. At this time, the W-type investment management data is less suitable as a segmentation feature when using the isolation forest algorithm to detect abnormal data in the Q-type investment management data.

[0049] By using the same calculation method for the degree of association between the Q-type investment management data and the W-type investment management data, the degree of association between any two types of investment management data in all types of investment management data can be obtained.

[0050] S3, taking the data at the same acquisition moment in any one type of investment management data and any remaining type of investment management data as a feature point pair; based on the occurrence frequency of the feature point pair and the time difference between the feature point pairs, determining the constraint factor between any one type of investment management data and any remaining type of investment management data.

[0051] Taking the Q-type investment management data and the W-type investment management data as an example, when analyzing the degree of association between the Q-type investment management data and the W-type investment management data, if both the Q-type investment management data and the W-type investment management data are data with relatively gentle fluctuations, but the degree of temporal association between the Q-type investment management data and the W-type investment management data is poor, it will cause a phenomenon of overestimation when calculating the degree of association between the Q-type investment management data and the W-type investment management data. Therefore, to more accurately reflect the degree of association between the Q-type investment management data and the W-type investment management data, in this embodiment, the data points at the same moment in the Q-type investment management data and the W-type investment management data are recorded as feature point pairs, and a statistical analysis is performed on the co-occurrence frequency of the obtained feature point pairs. The co-occurrence frequency is the ratio of the number of occurrences of each feature point pair in the Q-type investment management data and the W-type investment management data to the number of all feature point pairs, denoted as the third ratio, that is, the ratio of the number of the same feature point pairs to the number of all feature point pairs.

[0052] In this embodiment, a constraint factor is constructed through the distribution of the co-occurrence frequencies of each feature point pair. When both the Q-type investment management data and the W-type investment management data are relatively gentle in fluctuation, or even stable and unchanged, resulting in an overestimated degree of association between the Q-type investment management data and the W-type investment management data calculated, the number of feature point pairs statistically obtained is relatively small, and the co-occurrence frequency of each feature point pair is often large. For two types of investment management data with a relatively high actual degree of association, their data values are changing, and the obtained feature point pairs are relatively many, and the co-occurrence frequency of each feature point pair is often small.

[0053] Based on this, in this embodiment, a constraint factor between Q-type investment management data and W-type investment management data is constructed, and the specific calculation method is as follows: In the formula, y QW is the constraint factor between Q-type investment management data and W-type investment management data, M is the number of feature point pairs in Q-type investment management data and W-type investment management data, and f m is the co-occurrence frequency of the m-th feature point pair in Q-type investment management data and W-type investment management data, and LS m is the cumulative sum of the time intervals between all occurrences of the m-th feature point pair in Q-type investment management data and W-type investment management data.

[0054] It should be noted that feature point pairs with the same value at different times are regarded as one feature point pair. For example, if the set of feature point pairs between Q-type investment management data and W-type investment management data is [(1,2)(2,3)(4,6)(2,3)(5,7)(1,2)(2,3)], then the number of feature point pairs is 4, and the co-occurrence frequency of the feature point pair (1,2) is 2 / 4 = 0.5. If the times of each feature point pair in the set of feature point pairs [(1,2)(2,3)(4,6)(2,3)(5,7)(1,2)(2,3)] are 1, 2, 3, 4, 5, 6, 7 in sequence, then the cumulative sum of the time intervals between all occurrences of the feature point pair (2,3) is (4 - 2)+(7 - 4)=5.

[0055] y QW is the fusion result. Fusion refers to the way of combining multiple variables, and specifically, calculation methods such as addition, multiplication, addition-multiplication mixture, and taking the mean can be adopted.

[0056] It should be understood that the greater the co-occurrence frequency of each feature point pair, the more likely it is that the correlation degree between the calculated Q-type investment management data and W-type investment management data is on the high side. Therefore, the greater the constraint factor. When the cumulative sum of the time intervals between all occurrences of each feature point pair is greater, it indicates that the co-occurrence frequency distribution of the feature point pair is more discrete, and it is also more likely that both the Q-type investment management data and the W-type investment management data are relatively flat or stable, resulting in a relatively high correlation degree between the Q-type investment management data and the W-type investment management data, and the greater the constraint factor.

[0057] S4. Combine the correlation degree and the constraint factor to obtain the constraint correlation degree between any type of investment management data and the remaining types of investment management data; classify the remaining types of investment management data by using the correlation between the constraint correlation degrees; obtain the types of investment management data that have a strong correlation with any type of investment management data.

[0058] In this embodiment, the correlation degree between Q-type investment management data and W-type investment management data is constrained by a constraint factor, and the specific calculation method is as follows: In the formula, R QW is the constraint correlation degree between Q-type investment management data and W-type investment management data, GL QW is the correlation degree between Q-type investment management data and W-type investment management data, y QW is the constraint factor between Q-type investment management data and W-type investment management data, and μ is a preset value greater than 0 to avoid the denominator being 0. In this embodiment, μ = 0.01, and the implementer can set it according to the actual situation, which is not limited in this embodiment.

[0059] Using the same calculation method for the constraint correlation degree between Q-type investment management data and W-type investment management data, the constraint correlation degrees between Q-type investment management data and the remaining various types of investment management data are obtained. In this embodiment, the constraint correlation degrees between Q-type investment management data and all the remaining types of investment management data are clustered using the k-means clustering algorithm, and k = 2 is set. The implementer can set it according to the actual situation, which is not limited in this embodiment. Using the k-means clustering algorithm, the constraint correlation degrees between Q-type investment management data and all the remaining types of investment management data are divided into two categories, and the mean value of all the constraint correlation degrees in each category is calculated. The various types of investment management data in the category corresponding to the maximum mean value are used as the various types of investment management data that have a strong correlation with Q-type investment management data.

[0060] Among them, the k-means clustering algorithm is a well-known existing technology, and the implementer can select other existing feasible clustering algorithms by himself, which is not limited in this embodiment.

[0061] S5. In the various types of investment management data with strong correlation, the relevant data points of each data point in any one type of investment management data are obtained respectively. Using each data point and the relevant data point as the segmentation points respectively, the investment management data to which the segmentation points belong is segmented, and the number difference and the average level difference between the two categories of data after segmentation are analyzed. Combining the constraint correlation degree, the preference degree of each data point in any one type of investment management data as a segmentation point is evaluated.

[0062] When using the isolation forest algorithm to detect abnormal data in Q-type investment management data, the various types of investment management data that have a strong correlation with it are used as segmentation features to assist in the selection of the segmentation points of Q-type investment management data, so as to improve the segmentation efficiency and accuracy when constructing each node of the isolation tree. At the same time, it can also greatly reduce the depth of the isolation tree, thereby improving the efficiency and accuracy of abnormal data detection.

[0063] In this embodiment, the preference degrees of each data point in the Q-type investment management data as a segmentation point are calculated. Taking the data point q in the Q-type investment management data as an example, among various types of investment management data that have a strong correlation with the Q-type investment management data, the data points that are closest to the data point q in terms of collection time are respectively obtained as the relevant data points of the data point q in various types of investment management data that have a strong correlation with the Q-type investment management data. For various types of investment management data that have a strong correlation with the Q-type investment management data, using the relevant data points of various types of investment management data as the segmentation points, various types of investment management data are segmented into two parts. Specifically, in various types of investment management data, the data points greater than or equal to the relevant data points are used as one type of data points of various types of investment management data, and the data points less than the relevant data points are used as the second type of data points of various types of investment management data. Similarly, in the Q-type investment management data, the Q-type investment management data is segmented with the data point q as the segmentation point. The data points in the Q-type investment management data that are greater than or equal to the data point q are used as one type of data points, and the data points less than the data point q are used as the second type of data points.

[0064] Based on this, the preference degree of the data point q in the Q-type investment management data as a segmentation point is calculated. The specific calculation method is as follows: In the formula, Y Qq is the preference degree of the data point q in the Q-type investment management data as a segmentation point; n q1 is the number of one type of data points after the Q-type investment management data is segmented with the data point q as the segmentation point in the Q-type investment management data, and n q2 is the number of the second type of data points, is the mean value of one type of data points, is the mean value of the second type of data points, R u is the constraint correlation degree between the Q-type investment management data and the u-th type of investment management data that has a strong correlation with it; U is the number of types of investment management data that have a strong correlation with the Q-type investment management data, is the number of one type of data points after the u-th type of investment management data is segmented with the relevant data point of the data point q in the u-th type of investment management data, is the number of the second type of data points, is the mean value of one type of data points, is the mean value of the second type of data points.

[0065] It should be understood that when the investment type management data is segmented with the data point q and the relevant data point of the data point q, the greater the difference in the data volume and the average level between the two categories, the easier it is to find the abnormal data in the investment type management data when the data point q is used as the segmentation point. Therefore, the greater the preference degree of the data point q as a segmentation point, the more the data point q should be selected as the segmentation point.

[0066] S6. Use the isolation forest algorithm to construct isolation trees for any type of investment management data, select the splitting points in the construction process of the isolation trees based on the preference degree, obtain each piece of abnormal data in any type of investment management data and eliminate it, and perform investment management based on various types of investment management data after eliminating the abnormal data.

[0067] When using the isolation forest algorithm to detect abnormal data in Q-type investment management data, the first selected splitting point is the data point corresponding to the maximum preference degree in the Q-type investment management data. After splitting the Q-type investment management data, recalculate the preference degrees of each data point as splitting points in the split data sequence, and continue to split the data sequence using the maximum preference degree obtained from the recalculation until there is only one data point in the current subset of the isolation tree, or all data points in the subset are the same, or the isolation tree reaches the maximum depth. In this embodiment, the maximum depth is 8, and the implementer can set it according to the actual situation. This embodiment does not limit it here. Thus, the construction of the isolation tree in the isolation forest algorithm can be completed. Calculate the path length of the data points through the isolation tree, and then obtain the abnormal scores of the data points to complete the detection of abnormal data in the Q-type investment management data. Among them, the isolation forest algorithm is a well-known existing technology, and the specific process will not be elaborated. The flowchart of abnormal data detection is as Figure 2 shown.

[0068] Adopt the same abnormal data detection method as that for Q-type investment management data, obtain the abnormal data in various types of investment management data, eliminate the abnormal data in various types of investment management data, use the C4.5 decision tree construction algorithm for various types of investment management data after eliminating the abnormal data to complete the construction of the decision tree, and assist project investment decisions through the decision tree, thereby completing investment management based on the life cycle. Among them, the C4.5 decision tree construction algorithm is a well-known existing technology.

[0069] Based on the same inventive concept as the above method, the embodiment of the present application also provides an investment management system based on the life cycle, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-mentioned investment management methods based on the life cycle.

[0070] It should be noted that: the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specifically describes certain embodiments of this specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be beneficial.

[0071] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other, and the differences between each embodiment and other embodiments are emphasized.

[0072] The above are only the preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A life cycle-based investment management method, characterized in that: The method comprises the following steps: Collect various types of investment management data within a preset time period; Analyze the differences between adjacent data points in any type of investment management data, and the data distribution of the remaining types of investment management data within the collection time interval corresponding to the adjacent data points, to determine the correlation between the any type of investment management data and the remaining types of investment management data; Taking the data collected at the same time as any one of the investment management data and any other one of the investment management data as feature point pairs; determining the constraint factor between any one of the investment management data and any other one of the investment management data based on the frequency of occurrence of the feature point pairs and the time difference between the feature point pairs; Combining the correlation degree with the constraint factor, obtaining the constraint correlation degree between any one type of investment management data and the remaining types of investment management data; classifying the remaining types of investment management data using the correlation between the constraint correlation degrees; obtaining the types of investment management data that have a strong correlation with any one type of investment management data; Relevant data points of each data point in any one of the investment management data are obtained from the various types of investment management data with strong correlation, and the investment management data to which the segmentation points belong are segmented by taking the data points and the relevant data points as segmentation points, analyzing the difference in number of two types of data and the difference in average level between the various types of investment management data after segmentation, and evaluating the preference of each data point in any one of the investment management data as a segmentation point in combination with the constraint correlation degree; An isolation forest algorithm is used to construct an isolation tree for any type of investment management data, and a split point in the isolation tree construction process is selected based on the preference degree. Abnormal data in any type of investment management data is obtained and eliminated, and investment management is performed based on each type of investment management data after the abnormal data is eliminated.

2. The life cycle-based investment management method according to claim 1, characterized in that: The determination of the degree of association includes: Calculate the ratio of the absolute value of the difference between adjacent data points in any one type of investment management data to the maximum value among the adjacent data points, and record it as the first ratio; count the number of data points of the remaining types of investment management data that are not within the collection time interval of any one type of investment management data, and calculate the correlation degree by combining the first ratio with the number of data points and the data distribution of the remaining types of investment management data within the collection time interval corresponding to the adjacent data points in any one type of investment management data.

3. A life cycle-based investment management method as claimed in claim 2, characterized in that: The calculation method of the correlation degree is: In the formula, GL QW is the correlation between Q-type investment management data and W-type investment management data, exp[] is an exponential function with a natural constant as the base, N Q is the number of data points of Q-type investment management data, is the first ratio of the nth data point in the Q-type investment management data, obtain all the data points of the W-type investment management data in the collection time interval corresponding to the nth data point and the n+1th data point in the Q-type investment management data, calculate the ratio of the extreme value to the maximum value of the data points of all the W-type investment management data, record it as the second ratio, calculate the degree of dispersion of all the data points of the W-type investment management data, is the product of the second ratio and the discrete degree, β is a value preset to be greater than 0, and n W It is the number of data points in the W-type investment management data that are not within the collection time interval corresponding to the Q-type investment management data.

4. The life cycle-based investment management method according to claim 1, characterized in that: The determination of the constraint factor includes: Among all the feature point pairs of any one category of investment management data and any remaining category of investment management data, calculate the ratio of the number of occurrences of each feature point pair to the number of all feature point pairs, recorded as the third ratio, and calculate the cumulative sum of the time intervals between all occurrences of each feature point pair. The constraint factor is the fusion result of the third ratio and the cumulative sum.

5. The life cycle-based investment management method according to claim 1, characterized in that: The constraint association degree is positively correlated with the association degree and negatively correlated with the constraint factor.

6. The life cycle-based investment management method according to claim 1, characterized in that: The obtaining of various types of investment management data having a strong correlation with any type of investment management data includes: The constraint correlation between any one type of investment management data and all other types of investment management data is divided into two categories using a clustering algorithm, and the types of investment management data in the category with the largest mean constraint correlation are regarded as the types of investment management data with strong correlation to any one type of investment management data.

7. The life cycle-based investment management method according to claim 1, characterized in that: The relevant data points are data points in the various types of investment management data with strong correlation that are closest in collection time to the data points in any type of investment management data.

8. The life cycle-based investment management method according to claim 1, characterized in that: The preferred degree is calculated as follows: Where Y Qq is the preference of data point q as the split point in the Q-type investment management data; n q1 is the number of data points in one category after the Q-type investment management data is segmented with data point q as the segmentation point, n q2 is the number of data points of the second category, is the mean of one type of data points, is the mean of the two types of data points, R u is the constraint correlation degree between the Q-type investment management data and the u-th type of investment management data with strong correlation; U is the number of types of investment management data with strong correlation with the Q-type investment management data, is the number of data points in one category after the u-th category investment management data is segmented by the related data points of data point q, is the number of data points of the second category, is the mean of one type of data points, is the mean of the two types of data points.

9. The life cycle-based investment management method according to claim 8, characterized in that: When the Q-type investment management data is segmented using the data point q as the segmentation point, the data points in the Q-type investment management data that are greater than or equal to the data point q are regarded as the first-type data points, and the data points that are less than the data point q are regarded as the second-type data points; When the u-th type of investment management data is segmented by the data points related to data point q in the u-th type of investment management data, the same segmentation method as that used in the Q-type investment management data with data point q as the segmentation point is used for segmentation; The method of selecting the segmentation point of the isolated tree construction process based on the preference is: selecting the segmentation point with the largest preference as the segmentation point of the isolated tree construction process.

10. An investment management system based on life cycle, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 9 are implemented.

Images