CN109544346A - A kind of controllable type investment combination share-selecting method based on AP clustering algorithm - Google Patents

A kind of controllable type investment combination share-selecting method based on AP clustering algorithm Download PDF

Info

Publication number
CN109544346A
CN109544346A CN201811231235.6A CN201811231235A CN109544346A CN 109544346 A CN109544346 A CN 109544346A CN 201811231235 A CN201811231235 A CN 201811231235A CN 109544346 A CN109544346 A CN 109544346A
Authority
CN
China
Prior art keywords
cluster
stock
num
clustering algorithm
investment combination
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201811231235.6A
Other languages
Chinese (zh)
Inventor
程良伦
傅应龙
王卓薇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201811231235.6A priority Critical patent/CN109544346A/en
Publication of CN109544346A publication Critical patent/CN109544346A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/06—Asset management; Financial planning or analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Accounting & Taxation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Evolutionary Computation (AREA)
  • Game Theory and Decision Science (AREA)
  • Human Resources & Organizations (AREA)
  • Operations Research (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • General Business, Economics & Management (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

The invention discloses a kind of controllable type investment combination share-selecting method based on AP clustering algorithm, step includes: the stock pond data obtained in fixed time period, calculate earning rate of each branch stock in the stock pond obtained in the fixed time period, controllable parameter is set, the cluster of time series is carried out using AP clustering algorithm, calculate the similarity matrix of each stock in stock pond, Attraction Degree matrix and degree of membership matrix, obtain cluster result, and combine controllable parameter, judged, until cluster result meets controllable parameter, the cluster centre in each cluster is chosen as investment combination according to cluster result, carry out investment analysis.The present invention selects stocks by using the AP cluster of controllable type, greatly control in AP cluster that the cluster class number that occurs is excessive/very few to be led to not carry out correct Portfolio Selection Problem, the independence chosen in investment combination between each branch stock is improved simultaneously, reduces investment risk.

Description

A kind of controllable type investment combination share-selecting method based on AP clustering algorithm
Technical field
The present invention relates to investment tactics fields, more particularly, to a kind of controllable type investment group based on AP clustering algorithm Close share-selecting method.
Background technique
Investment way of the investor by assets investment in two or more financial product is known as investment combination.It throws Providing Combination thought is that the wealth of investor is dispersed into more parts, the investment risk of investor is reduced, so that investor is not because of market Fluctuation greatly brings unnecessary loss.And during the selection of investment combination, since the quantity of personal share is big, stock market is related Property be affected, optimal investment combination candidate's stock is not easily found during stock selection, therefore in reality Cannot evade in investment combination influences loss of assets brought by the risk of multiply due to one, is huge for investor Challenge.
It is directed to investment combination to select stocks, many researchers are directed to investment combination research in recent years, they mainly pass through pair The correlation of assets is analyzed, to obtain the less investment combination of each stock correlation.South China Science & Engineering University Sun Wei et al. is mentioned A kind of fuzzy stochastic investment combination method is developed in fuzzy stochastic mean-standard deviation method out, solves different type investor's Portfolio Selection Problem；Hunan University Wu Hui etc. proposes the theory converted based on macroeconomy, it is found that it has Portfolio Selection Significant impact；Li Jianfu etc. studies Index Portfolio Selection weight mechanism select permeability, and discovery basic side method of weighting is being thrown Behave oneself best in money combination.Research in recent years is based substantially on the conventional learning algorithms of traditional investment combination, by investment The preference of person and relevant research of selecting stocks carried out to the fluctuation in market, but there is no the investment combination of stock selected by considering whether Meet real investment value, and whether there is the ability for reducing risk；And research process is based on traditional statistical method, and The machine learning bring method in reality is not accounted for, therefore investment combination research is also in an elementary step, Wu Fazhen Just it is applied to market.Select stocks and whether meet the relativity problem of investment combination in the process, and corresponding to selecting stocks during be It is no reach reduction select stocks risk the problems such as, be investment combination be worth research direction.
Summary of the invention
The present invention in order to overcome at least one of the drawbacks of the prior art described above, provide it is a kind of based on AP clustering algorithm can Control formula investment combination share-selecting method.
The present invention is directed to solve above-mentioned technical problem at least to a certain extent.
Primary and foremost purpose of the invention is to provide a kind of controllable type investment combination share-selecting method based on AP clustering algorithm,
In order to solve the above technical problems, technical scheme is as follows:
A kind of controllable type investment combination share-selecting method based on AP clustering algorithm, the method includes following processing steps:
S1: the stock pond data in fixed time period are obtained；
S2: earning rate of each branch stock in the fixed time period in the stock pond of acquisition is calculated；
S3: setting controllable parameter；
S4: carrying out the cluster of time series using AP clustering algorithm, calculates the similarity matrix of each stock in stock pond, inhales Degree of drawing matrix and degree of membership matrix obtain cluster result, and combine controllable parameter, are judged, until cluster result meets Until controllable parameter；
S5: the cluster centre in each cluster is chosen as investment combination according to cluster result, carries out investment analysis.
Further, the fixed time period can customize setting, and the unit of period is day, and minimum time section is 30 It.
Further, each branch stock yield formula calculated in stock pond is as follows: P_value=(closei- closei-1)/closei-1, wherein P_value indicates stock yield, represents increasing of the same day income compared to the income of yesterday Long rate, closei represent stock same day closing price, and closei-1 represents stock proxima luce (prox. luc) closing price, and each branch stock is with earning rate Based on the time series within the fixed section time.
Further, the controllable parameter includes number clus_num_max in maximum cluster, number clus_num_ in most tuftlet Min, ideal number of clusters cluster_num；Number clus_num_max is used for when number is higher than in the cluster in cluster process in maximum cluster Upper limit threshold numhWhen, carry out segmentation in subsequent cluster；Number clus_num_min is used for when the cluster in cluster process in most tuftlet Interior number is lower than lower threshold numl, and when number in most tuftlet is not satisfied in the number in each cluster, carry out between subsequent cluster Merge；Ideal number of clusters cluster_num is used for the asset allocation share for investor to investment and sets up, to meet investment The number of the desired separated investment assets of person.
Further, the upper limit threshold numh, lower threshold numlIt is determined according to different number of clusters k, k gets N, N from 1 The sum for indicating stock records after carrying out k-means cluster in the case where different k average departure centre distance in class, and passes through The mapping of elbow method judges number of clusters of the k corresponding to the maximum point of slope variation as cluster, calculating upper limit threshold value numh=N/ (k- 1), lower threshold numl=N/ (k+1).
Further, the cluster of time series, each branch stock energy in the stock pond of acquisition are carried out using AP clustering algorithm It is enough as being the time series based on earning rate, or as the time series based on earning rate variation tendency.It is dimension with number of days Degree, calculates the similarity matrix of each branch stock, obtains Attraction Degree matrix by similarity matrix and degree of membership matrix, update change Generation, and controllable parameter is combined, carry out the cluster in stock pond.
Further, after carrying out AP cluster, candidate stock of the cluster centre as investment combination in each cluster is chosen.
Compared with prior art, the beneficial effect of technical solution of the present invention is:
The present invention selects stocks by using the AP cluster of controllable type, greatly controls the cluster class number occurred in AP cluster Mesh is excessive/very few to be led to not carry out correct Portfolio Selection Problem, while being improved and being chosen in investment combination between each branch stock Independence, reduce investment risk.
Detailed description of the invention
Fig. 1 is algorithm flow chart.
Fig. 2 is the daily earning rate information of 300 constituent stocks of Shanghai and Shenzhen wherein five stock in certain time.
Fig. 3 is that AP clusters similarity matrix.
Fig. 4 is stock principle of similitude figure.
Fig. 5 is that controllable type AP clusters flow chart a.
Fig. 6 is that controllable type AP clusters flow chart b.
Fig. 7 is that controllable type AP clusters flow chart c.
Specific embodiment
The following further describes the technical solution of the present invention with reference to the accompanying drawings and examples.
Embodiment 1
A kind of controllable type investment combination share-selecting method based on AP clustering algorithm, step include:
1. the stock pond for crawling in the fixed section time and (recommending 30≤day≤50) such as 300 constituent stocks of Shanghai and Shenzhen, upper card refer to Number, Shenzhen Stock Exchange are used as stock pond at finger etc..The dimension that number of days is too long to will lead to cluster calculates excessive, and too short meeting of number of days is so that cluster Accuracy is not high, and the set time is with day unit, the range of recommendation 30 to 50 days.
2. calculating each branch stock yield in stock pond, calculation formula is as follows:
P_value=(closei-closei-1)/closei-1,
Wherein, P_value indicates stock yield, represents growth rate of the same day income compared to the income of yesterday, therefore Each branch stock is based on the time series within certain time of earning rate.The earning rate P_ of 5 stocks is had recorded in Fig. 2 The information of value, wherein have stock1, stock2, stock3, stock4, stock5, five stocks, △ (stocki, Stockj), expression be stocki and stockj two stock earning rate difference, as shown be each of five stocks with The earning rate difference of other stocks, from some date date (i), it can be seen that for stock1, stock2 compared to The earning rate of stock3 is closer, and other periods are also in this way, therefore, one can consider that stock2 is phase from trend It is similar to stock1, there is certain correlation.In stock3, stock4, stock5, stock4 and stock5 are in each time The sum of yield gap in section is smaller than stock3, therefore one can consider that within the specific period, stock4 with Stock5 has higher correlation.Therefore, by judging the deviation of the earning rate in certain time, it can be seen that its correlation It influences, relativity measurement index is as follows:
P_value (stocka) indicates the earning rate of stock a, and P_value (stockb) indicates the earning rate of stock b The distance of stock a and stock b in period date (i) to date (j) are indicated in Relation (stocka, stockb) formula Difference, it can be understood as the Euclidean distance based on trend.In addition, other daily indexs, such as simple moving average also can be selected The indexs such as SMA, weighted moving average WMA carry out clustering, and selection earning rate carries out analysis and is derived from earning rate more here It can reflect the Long-term change trend of a stock, and calculate simply, be easy to get.
3.AP clustering algorithm carries out the cluster of time series, and each branch stock in the stock pond of acquisition is considered as to be based on The time series of earning rate can also see the time series based on earning rate variation tendency as.Using number of days as dimension, each is calculated The similarity matrix of stock, as shown in Figure 3.Similarity is regarded as ability of the stock A as the cluster centre of stock B, here Using negative Euclidean distance, as shown in figure 4, if the value of S is bigger, similarity is higher, with stock stock1 and stock stock2 For, it can more illustrate stock stock1Ability as cluster centre is stronger.Furthermore Attraction Degree square is obtained by similarity matrix Battle array and degree of membership matrix update iteration, and combine controllable parameter, carry out the cluster to stock pond.Attraction Degree matrix and return The formula of category degree matrix is as follows:
Attraction Degree matrix calculates are as follows: Rt+1(i, k)=(1- λ) Rt+1(i,k)+λ·Rt(i,k)
Wherein, Attraction Degree matrix is updated to
λ in formula is learning rate, and value range [0.5,1), the convergence for algorithm.
Wherein Rt+1(i, k) indicate new point i for the Attraction Degree of point k, by i for k similarity S (i, k) and i for The degree of membership A of other points (not including k)t(i, j) and the old Attraction Degree R for other pointst(i, j) is determined；Work as i When with k being same, i.e., it need to be only compared by original similarity with the similarity of other points, choose maximum be used as more Newly.The sum of angelica degree and Attraction Degree are bigger, then can illustrate that k is bigger as the probability of cluster centre.
Degree of membership matrix calculates are as follows: At+1(i, k)=(1- λ) At+1(i,k)+λ·At(i,k)
Wherein, degree of membership matrix update is
λ in formula is learning rate, and value range [0.5,1), the convergence for algorithm.
Wherein At+1(i, k) is new degree of membership matrix, indicate i belong to point k degree be have it is much, by comparing more The Attraction Degree R of point k after newt+1Whether the size of the Attraction Degree of (i, k) and other points, decision-point i should be attributed to centered on point k Class.
Attraction Degree matrix puts the ability for other each points as cluster centre, if it is bigger, the point to analyze certain It can be used as a cluster centre；The ability which central point degree of membership matrix should be attributed to illustrate the point, angelica degree are got over Height, then the point should be attributed in the corresponding highest point of degree of membership by explanation；One point of comprehensive descision belongs to center, is still attributed to Some point, takes maximization to compare, to carry out the classification to point.The number in cluster is judged as shown in Fig. 5,6,7, after classification Whether mesh meets set controllable factor, and separation or merging in cluster are carried out in the case of being unsatisfactory for, and obtains stock cluster to the end As a result.
The terms describing the positional relationship in the drawings are only for illustration, should not be understood as the limitation to this patent；
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this Made any modifications, equivalent replacements, and improvements etc., should be included in the claims in the present invention within the spirit and principle of invention Protection scope within.

Claims (7)

1. a kind of controllable type investment combination share-selecting method based on AP clustering algorithm, which is characterized in that the method includes following Processing step:
S1: the stock pond data in fixed time period are obtained；
S2: earning rate of each branch stock in the fixed time period in the stock pond of acquisition is calculated；
S3: setting controllable parameter；
S4: carrying out the cluster of time series using AP clustering algorithm, calculates the similarity matrix of each stock, Attraction Degree in stock pond Matrix and degree of membership matrix obtain cluster result, and combine controllable parameter, are judged, until cluster result meets controllably Until parameter；
S5: the cluster centre in each cluster is chosen as investment combination according to cluster result, carries out investment analysis.
2. a kind of controllable type investment combination share-selecting method based on AP clustering algorithm according to claim 1, feature exist In the fixed time period can customize setting, and the unit of period is day, and minimum time section is 30 days.
3. a kind of controllable type investment combination share-selecting method based on AP clustering algorithm according to claim 1, feature exist In each branch stock yield formula calculated in stock pond is as follows: P_value=(closei-closei-1)/closei-1, In, P_value indicates stock yield, represents same day growth rate of the income compared to the income of yesterday, closeiRepresent stock Same day closing price, closei-1Stock proxima luce (prox. luc) closing price is represented, each branch stock is based on earning rate in the fixed section time Interior time series.
4. a kind of controllable type investment combination share-selecting method based on AP clustering algorithm according to claim 1, feature exist In the controllable parameter includes number clus_num_max in maximum cluster, number clus_num_min, ideal number of clusters in most tuftlet cluster_num；Number clus_num_max is used for when number is higher than upper limit threshold num in the cluster in cluster process in maximum clusterh When, carry out segmentation in subsequent cluster；Number clus_num_min is used under number is lower than in the cluster in cluster process in most tuftlet Limit threshold value numl, and when number in most tuftlet is not satisfied in the number in each cluster, merge between subsequent cluster；Ideal number of clusters Cluster_num is used for the asset allocation share for investor to investment and sets up, for the desired separated for meeting investor The number of investment assets.
5. a kind of controllable type investment combination share-selecting method based on AP clustering algorithm according to claim 4, feature exist In the upper limit threshold numh, lower threshold numlIt is determined according to number of clusters k, k gets N from 1, and N indicates the sum of stock, record Average departure centre distance in class is carried out after k-means cluster in the case where different k, and its slope is judged by the mapping of elbow method Change k corresponding to maximum point, as the number of clusters of cluster, calculating upper limit threshold value numh=N/ (k-1), lower threshold numl= N/(k+1)。
6. a kind of controllable type investment combination share-selecting method based on AP clustering algorithm according to claim 1, feature exist In using the cluster of AP clustering algorithm progress time series, it is based on receipts that each branch stock in the stock pond of acquisition, which can be used as, The time series of beneficial rate, or as the time series based on earning rate variation tendency.Using number of days as dimension, each branch stock is calculated Similarity matrix, Attraction Degree matrix and degree of membership matrix are obtained by similarity matrix, update iteration, and combines controllable ginseng Number carries out the cluster in stock pond.
7. a kind of controllable type investment combination share-selecting method based on AP clustering algorithm according to claim 1, feature exist In, carry out AP cluster after, choose candidate stock of the cluster centre as investment combination in each cluster.
CN201811231235.6A 2018-10-22 2018-10-22 A kind of controllable type investment combination share-selecting method based on AP clustering algorithm Pending CN109544346A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811231235.6A CN109544346A (en) 2018-10-22 2018-10-22 A kind of controllable type investment combination share-selecting method based on AP clustering algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811231235.6A CN109544346A (en) 2018-10-22 2018-10-22 A kind of controllable type investment combination share-selecting method based on AP clustering algorithm

Publications (1)

Publication Number Publication Date
CN109544346A true CN109544346A (en) 2019-03-29

Family

ID=65844162

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811231235.6A Pending CN109544346A (en) 2018-10-22 2018-10-22 A kind of controllable type investment combination share-selecting method based on AP clustering algorithm

Country Status (1)

Country Link
CN (1) CN109544346A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110362558A (en) * 2019-06-12 2019-10-22 广东工业大学 A kind of energy consumption data cleaning method based on neighborhood propagation clustering
CN111179077A (en) * 2019-12-19 2020-05-19 成都数联铭品科技有限公司 Method and system for identifying abnormal stock transaction

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110362558A (en) * 2019-06-12 2019-10-22 广东工业大学 A kind of energy consumption data cleaning method based on neighborhood propagation clustering
CN110362558B (en) * 2019-06-12 2022-12-16 广东工业大学 Energy consumption data cleaning method based on neighborhood propagation clustering
CN111179077A (en) * 2019-12-19 2020-05-19 成都数联铭品科技有限公司 Method and system for identifying abnormal stock transaction
CN111179077B (en) * 2019-12-19 2023-09-12 成都数联铭品科技有限公司 Stock abnormal transaction identification method and system

Similar Documents

Publication Publication Date Title
Li et al. An overview of personal credit scoring: techniques and future work
Balakrishnan et al. A study of the classification capabilities of neural networks using unsupervised learning: A comparison with K-means clustering
Nguyen et al. Learning pattern classification tasks with imbalanced data sets
Zhu et al. Particle Swarm Optimization (PSO) for the constrained portfolio optimization problem
Witten et al. Supervised multidimensional scaling for visualization, classification, and bipartite ranking
CN109034231A (en) The deficiency of data fuzzy clustering method of information feedback RBF network valuation
CN110210973A (en) Insider trading recognition methods based on random forest and model-naive Bayesian
CN109390037A (en) The full site recognition methods of mature miRNA based on SVM-AdaBoost
CN109544346A (en) A kind of controllable type investment combination share-selecting method based on AP clustering algorithm
Liang et al. An Adaptive News-Driven Method for CVaR-sensitive Online Portfolio Selection in Non-Stationary Financial Markets.
CN109948534A (en) The method for carrying out recognition of face is clustered using fast density peak value
Damoulas et al. Inferring sparse kernel combinations and relevance vectors: an application to subcellular localization of proteins
Kadam et al. Machine learning methode for automatic potato disease detection
CN117253550A (en) Spatial transcriptome data clustering method
Khan et al. DeepGene transformer: Transformer for the gene expression-based classification of cancer subtypes
CN111914900B (en) User electricity utilization mode classification method
CN113688960A (en) Grey wolf optimization GHFCM-based residential power data clustering method and device
Krawczyk et al. Breast thermogram analysis using a cost-sensitive multiple classifier system
Ahmadi et al. Model order selection for multiple cooperative swarms clustering using stability analysis
CN107909478A (en) FOF mutual fund portfolio system and methods based on social network clustering and information gain entropy index
CN106354886A (en) Method for screening nearest neighbor by using potential neighbor relation graph in recommendation system
Chang et al. Stock portfolio construction using evolved bat algorithm
Li et al. A fuzzy linear programming-based classification method
Bhat et al. DeepCancer: Detecting cancer via deep generative learning through gene expressions
Wan et al. Sddsmote: Synthetic minority oversampling technique based on sample density distribution for enhanced classification on imbalanced microarray data

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20190329