CN112612942B - Social big data-based fund recommendation system and method - Google Patents

Social big data-based fund recommendation system and method Download PDF

Info

Publication number
CN112612942B
CN112612942B CN202011596610.4A CN202011596610A CN112612942B CN 112612942 B CN112612942 B CN 112612942B CN 202011596610 A CN202011596610 A CN 202011596610A CN 112612942 B CN112612942 B CN 112612942B
Authority
CN
China
Prior art keywords
fund
user
influence
node
social
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.)
Active
Application number
CN202011596610.4A
Other languages
Chinese (zh)
Other versions
CN112612942A (en
Inventor
陶飞飞
石敏芳
谭梦婕
王雅淳
庄展鹏
刘生伟
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.)
Hohai University HHU
Original Assignee
Hohai University HHU
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 Hohai University HHU filed Critical Hohai University HHU
Priority to CN202011596610.4A priority Critical patent/CN112612942B/en
Publication of CN112612942A publication Critical patent/CN112612942A/en
Application granted granted Critical
Publication of CN112612942B publication Critical patent/CN112612942B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/951—Indexing; Web crawling techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9536—Search customisation based on social or collaborative filtering
    • 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
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0631—Item recommendations
    • 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
    • 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
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/01—Social networking
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Physics & Mathematics (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • General Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Human Resources & Organizations (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Health & Medical Sciences (AREA)
  • Tourism & Hospitality (AREA)
  • Computing Systems (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Technology Law (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

The invention discloses a fund recommendation system and method based on social big data. The financial social big data acquisition module is used for crawling behavior and public opinion data of users of the financial social network site, the fund popularity analysis module is used for conducting popularity analysis on popular fund products of the financial social network site, and the fund net worth risk analysis module judges fund investment risk according to the current market average market profitability value. The recommendation system can better serve investor groups with incomplete investment and financing bases and experiences, and the recommendation method analyzes user behaviors and fund popularity on a financial social network site by designing a self-adaptive user influence algorithm, and recommends fund products suitable for investment for clients and social circle friends by combining relevant factors such as fund evaluation values and market risks.

Description

Social big data-based fund recommendation system and method
Technical Field
The invention relates to a fund recommendation system and method, in particular to a fund recommendation system and method based on social big data.
Background
Along with the gradual popularization of the public internet financing concept, the financing scale is enlarged, the characteristics of low cost, low threshold, mass customized financing scheme and the like of intelligent investment meet the financing requirements of the middle-yield level and the mass abundance level, the threshold of investment advisor service is reduced, and common financing users can enjoy the investment advisor service. However, most investors lack systematic and scientific investment concepts, the investment behavior is driven by products more, and many people have not yet contacted a recommendation system about fund investment.
The effective analysis of financial investment products and the recommendation of proper investment products to users are always the targets expected by experts and practitioners in the field of financial investment. In the age of rapid development of the internet at present, any information and public events are possibly amplified, and in various financial social websites, public opinion information about investment products such as stocks and funds is dazzling to common investors and rational and objective choices are difficult to make. At present, the mainstream intelligent investment model and algorithm are homogenized, the risk preference evaluation of the user is very simple, and the financing requirement of an investor is presumed only through simpler questions and answers, so that the investment product which is possibly recommended to the investor and does not really meet the fund investment product of the user is possibly recommended.
Disclosure of Invention
The invention aims to: the invention aims to provide a fund recommendation system based on social big data, which realizes judgment and guidance of fund investment, avoids risks to the greatest extent and improves client investment income.
The technical scheme is as follows: the fund recommendation system based on the social big data comprises a financial social big data acquisition module, a fund heat degree analysis module, a fund net worth risk analysis module and a fund screening and recommending module;
the financial social big data acquisition module is used for acquiring and processing investment preference information of the user and crawling behavior and public opinion data of the user of the financial social website;
the fund popularity analysis module is used for constructing a dynamic social topological network structure, designing a self-adaptive user influence strength algorithm and carrying out popularity analysis on popular fund products in the financial social network site;
the fund net value risk analysis module is used for judging the risk of fund investment according to the comparison of the average market profit value of the current market and historical high-level and low-level numerical values;
and the fund screening and recommending module is used for integrating results of fund evaluation, market risk analysis and fund popularity analysis, and screening and recommending fund products meeting the requirements of customers and friends in social circles of the customers.
In conclusion, the financial social big data acquisition module provides data storage for the recommended net fund analysis and recommendation according to the emotional tendency of the user of the social network; the fund popularity analysis module introduces the self-adaptive weight factor, effectively realizes the analysis of the investment preference, the behavior and the public opinion of the user, and recommends the fund suitable for the investment preference of the client through popularity analysis on the popular fund of the financial social network site, thereby realizing the income maximization of the investor.
The fund recommendation method based on the social big data comprises the following steps:
(1) collecting and processing investment preference information of a user and crawling behavior and public opinion data of a user of a financial social network site;
(2) constructing a dynamic social topological network structure, designing a self-adaptive user influence strength algorithm, and carrying out heat analysis on popular fund products in the financial social network site;
(3) judging the risk of fund investment according to the comparison of the average market profitability value of the current market with historical high-level and low-level numerical values;
(4) and (4) integrating the results of fund valuation, market risk analysis and fund popularity analysis, and screening and recommending fund products meeting the requirements of customers and friends in social circles of the customers.
Further, in the step (1), a Hadoop-based big data platform is built to store the investment preference information of the user and the behavior and public opinion data of the user, and a MapReduce frame is adopted to clean and preprocess the financial social big data.
In the step (2), a dynamic social topological network structure is constructed, self-adaptive weight factors are introduced, a self-adaptive user influence strength algorithm is designed, and the transfer probability from a user node of a social network site to a neighbor node of the user node is automatically distributed, so that the transfer between the nodes has certain self-adaptive characteristics, and the heat analysis is carried out on popular fund products on the social network site.
Preferably, the adaptive weight factors include inter-node influence, user behavior influence and fund heat topic influence.
In the step (2), the popularity analysis is combined with the emotion index of the netizen of the financial social network site; the self-adaptive user influence degree algorithm selects the forwarding number, the comment number, the praise number, the fan number and whether the forwarding number, the comment number, the praise number and the fan number are opinion leader parameters, so that more optimal distribution of the transfer probability from the social user node to the neighbor node is realized, the node propagation capacity and the accuracy of the user influence degree measurement are effectively improved, and the heat degree of the fund can be analyzed more accurately.
And (3) acquiring real-time net worth estimation, unit net worth and accumulated net worth of the fund products, performing incremental crawling operation every day, and judging the risk of fund investment according to comparison of the average market profitability value of the current market with historical high-level and low-level numerical values.
In step (3), a script framework based on python collects real-time net value estimation, unit net value and accumulated net value of the fund.
In the step (3), the judgment standard of the risk of fund investment is as follows: when the average market profitability value of the market is below 10 times, the market estimation value belongs to the underestimation category; the market average market profitability value is within the range of 10-15 times, and the market average market profitability value belongs to the reasonable evaluation category; when the average market profitability value of the market is more than 15 times, the long-term investment value of the current market is reduced; and when the market average market profitability value reaches more than 20 times, the market average market profitability value belongs to the category of overestimation, and the relative risk is judged according to the interval where the market profitability value is located.
In the step (4), recommending the investment scheme meeting the requirements of the client to the client, and if the investment scheme does not meet the requirements of the client, repeating the steps (2) and (3) to adjust the combination; the fund product recommended for the client also supports the purchase of the fund directly on the existing scheme after the friends in the social circle enter the platform in a link sharing mode.
In the step (4), a recommendation scheme after user investment preference analysis is performed according to the questionnaire can be directly provided, and meanwhile, a link shared by friends in a social circle is provided to enter a platform, and fund investment is performed by directly referring to the scheme of the friends. As most of users in the same social circle have similar economic abilities, the fund investment scheme of friends has higher recommendation reference value, and the function can facilitate users without or with weak financial bases to more conveniently select fund investment.
Has the beneficial effects that: compared with the prior art, the invention has the following remarkable advantages: by analyzing the social big data, optimizing a user influence strength algorithm, constructing a dynamic social topological network structure, obtaining popular fund products of a financial social circle, comprehensively judging the risk of fund investment according to the comparison of the average market profit value of the current market and historical high-order and low-order numerical values, and finally forming an effective fund recommendation scheme.
Drawings
FIG. 1 is a system block diagram of a social big data based fund recommendation system;
FIG. 2 is a flow diagram of the operation of a social big data based fund recommendation system;
FIG. 3 is a diagram of a social network topology incorporating context nodes;
FIG. 4 is a schematic diagram of a user influence strength algorithm based on an adaptive factor.
Detailed Description
The technical solution of the present invention is further illustrated by the following examples.
The core of the invention is a fund recommendation system based on social big data, which is convenient for serving investors with weak investment and financing bases and provides a better scheme for the investors in fund investment selection. As shown in FIG. 1, the fund recommendation system based on social big data comprises a financial social big data acquisition module, a fund heat analysis module, a fund net worth risk analysis module and a fund screening recommendation module.
The financial social big data acquisition module is used for acquiring investment preference information of users and behavior and public opinion data of users crawling financial social network sites, building a big data platform storage data based on Hadoop, and cleaning and preprocessing the data by adopting a MapReduce frame;
the fund popularity analysis module constructs a dynamic social network topological structure, introduces self-adaptive weight factors (influence among nodes, user behavior influence and fund popularity topic influence), designs a self-adaptive user influence strength algorithm and carries out popularity analysis on popular funds on the financial social network site;
the fund net value risk analysis module is used for acquiring real-time net value estimation, unit net value and accumulated net value of the fund, performing incremental crawling operation every day, and judging the risk of fund investment according to the comparison of the average market profit value of the current market with historical high-level and low-level numerical values;
and the fund screening and recommending module is used for recommending appropriate fund investment products for the customers according to comprehensive screening results of fund equity risk analysis and fund popularity analysis, and can further recommend own fund purchasing schemes to be shared with friends in the social circle.
FIG. 2 is a flow chart of an embodiment of the present invention, which includes the following steps:
step 1: relevant information of the fund, basic information (including user investment preference information) in a questionnaire of the client A and user behavior data on a financial social network site are collected.
The relevant information of the fund includes not only real-time net value estimation, unit net value and accumulated net value of the fund, but also information of the fund, such as risk degree, fund size and the like, which is beneficial to the user. After the data are added to the big data storage system, the data can be updated according to the actual market conditions and the changes of the financial social network sites. The behavior data of the user comes from a snow net, an east wealth net and the like, and comprises a series of behavior messages of user posting, praise, forwarding and the like. Useful emotion evaluation information is extracted and stored in an HDFS file system at the bottom layer of Hadoop.
The big data system can adopt Hadoop, and the Hadoop is composed of components such as HDFS, MapReduce, HBase, Hive and ZooKeeper. Data collection is mainly based on the crawling framework crawling of python, and data updating crawling operation is carried out every day so as to guarantee public opinion data based on the current changed fund market and the financial social network.
Step 2: by utilizing factors such as influence among nodes, influence of user behaviors, influence of fund popularity topics and the like, a self-adaptive influence strength algorithm is designed, the user behaviors and public opinion data of the social network site are analyzed, the popularity of fund products is obtained, and the popular fund investment products of the current social network site are given.
In a specific implementation, the step may utilize a MapReduce distributed computing framework. The parallel processing model has the characteristics of high expandability, convenience in application and the like, and the parallelization of the traditional MR _ LSH algorithm is improved on the basis of MapReduce, so that the efficient clustering of large-scale graph data can be realized in a distributed cluster environment.
And step 3: and comprehensively judging whether the fund investment is suitable for being carried out at present according to risk factors such as fund evaluation values, market profitability values and the like, and adjusting a recommendation scheme according to a judgment result. Wherein, the market risk judgment standard of fund investment is as follows: when the average market profitability value of the market is less than 10 times, the market estimation value belongs to the underestimation category; the market average market profitability value is within the range of 10-15 times, and the market average market profitability value belongs to the reasonable evaluation category; when the average market profitability value of the market is more than 15 times, the long-term investment value of the current market is reduced; and when the market average market profitability value reaches more than 20 times, the market average profitability value belongs to the overestimation category, and the relative risk is judged according to the interval where the market profitability value is located.
During specific implementation, secondary analysis is performed on the recommendation result under the basic conditions of combining the economic capability of the client, risk preference and the like, and when client data are analyzed, the system effectively classifies client groups by adopting a connected graph clustering algorithm based on GraphX, so that the accuracy of the recommendation result is improved.
And 4, step 4: and recommending the investment scheme of the fund to the user A, wherein the user can share the investment scheme of the fund to the user B through a social circle.
In specific implementation, the step can be mainly based on mobile-side wechat push or an applet, and further recommend and share the fund purchase scheme of the customer to the social circle friends, for example, a forwarding function is implemented by adopting an onshareepromessage (option) function in Page in the wechat applet.
In step 2, as shown in fig. 3, a background Node "is added to the social user network topology structure, and the" group Node "and all nodes in the graph are interconnected in a bidirectional manner, so that the original network G (N, M) with N nodes and M edges is converted into a strongly connected network with N +1 nodes and M +2N edges. The newly introduced nodes enable the graph transformation to be completely communicated, and meanwhile, the problem that the nodes influence the convergence and the jump parameters of the strength algorithm is solved.
After introducing the group Node, the iterative formula of the Node influence is shown as the following graph:
Figure BDA0002868343540000051
in the above formula, i, j is 1,2, …, N + 1. If node i is directly connected to node j, then a ij 1, otherwise 0,
Figure BDA0002868343540000052
which represents the out-degree of the node j,
Figure BDA0002868343540000053
representing the probability that node i will randomly walk to node j. All nodes except the background node have LR as initial value i The background node g is initialized to LR at 1 g 0. The LR value at node i in steady state is:
Figure BDA0002868343540000054
in the above formula, t c Indicating the number of convergence, LR g (t c ) Is the LR value of node g in steady state. After iteration is finished, each Node obtains a stable influence value, and because the influence ranking is not included in the 'group Node', the converged influence value is averagely distributed to other nodes.
From this, an Adaptive User Influence metric Algorithm (AUIR) is further proposed, as shown in fig. 4:
(1) the probability of each node jumping to the neighbor node is equal, and each node cannot self-adaptively set the jumping probability according to the attribute behavior of the actual neighbor node. Aiming at the problem, the AUIR algorithm introduces a self-adaptive weight factor, and automatically distributes the transition probability from the user node to the neighbor nodes thereof, so that the inter-node transition has certain self-adaptive characteristics.
(2) The AUIR algorithm calculates the weight occupied by each Node based on the self-adaptive weight factor, and differentially distributes the value of 'group Node', thereby measuring the user influence value more accurately.
In the process of providing the self-adaptive weight factors, the algorithm considers three factors, namely the influence among the nodes, the influence of user behavior and the influence of fund heat topic by analyzing the node attribute behaviors.
(1) Influence between nodes
In a directed social network, the out-degree of a node represents the interest of the node, and the in-degree represents the endorsement of other nodes to the node. If the number of the nodes pointed by the two nodes is more, the interests of the two nodes are more similar; if the number of nodes which point to two nodes together is larger, the acceptance of the two nodes is higher, and the interests of the two nodes are similar. Therefore, if the number of common neighbor nodes between two nodes is more, that is, the number of nodes pointed to by the two nodes is more, or the number of nodes pointed to by the two nodes is more, the more similar the two nodes are, the higher the transition probability between the nodes is.
In an undirected graph, the similarity of two nodes is measured by a Jaccard similarity coefficient, and if the Jaccard value is larger, the similarity is larger, and the formula is as follows:
Figure BDA0002868343540000061
considering that in the directed graph, the node similarity is respectively related to the number of nodes pointed to by the node similarity and the number of nodes pointed to by the node similarity, the Jaccard metric formula is improved as shown in the following formula:
Figure BDA0002868343540000062
in the above formula, w n (i, j) represents the influence between nodes, the value of which is between 0 and 1, i and j are nodes in the graph respectively, n (i) in And n (j) in The number of nodes pointing to i and j, respectively, n (i) out And n (j) out I and j respectively refer to the number of nodes pointed by i and j together.
(2) Influence of user behavior
In a social network, user behavior attributes influencing the influence of a user are more, and five representative attributes are selected in the section and are respectively: number of forwards, number of comments, number of praise, number of fan, and whether or not to comment leader. The attribute list and the identifier thereof are specifically listed in table 1:
TABLE 1 user behavior Attribute List
Figure BDA0002868343540000071
Here, the user behavior attributes listed in the above table are quantized, and a user behavior influence strength formula shown as follows is provided, specifically, the formula is shown as formula (5):
Figure BDA0002868343540000072
in the above formula, A i The user i is the beating frequency in the statistical period, namely the activity of the user i in the statistical period. Wherein M is i The number of beats of the user i in the T period is shown, and T is the statistical period. T is i Represents the propagation capability of user i in a specified period, where m j Indicating any blog sent by user i during period T,
Figure BDA0002868343540000073
indicating Bowen m j The forward-to-forward rate of (c),
Figure BDA0002868343540000074
are respectively m j Comment rate and like rate. T is i The average number of praise, the average number of comments and the average number of forwarding of the user i in the specified time period T are included. The higher the user liveness, the stronger the post propagation capacity, and the greater the impact of the user. M is the fan value of the user with the most fans in the users who pay attention to i, V (i) represents whether the i user is a large V user, if the i user is a large V user, the I user is 1, and if the I user is not a large V user, the I user is 0. a. b and c are respective weights.
(3) Influence of heat topic of fund
The users of the social network sites have topic bias, that is, if the similarity between the blogged text sent by the users and the topic related to the fund popularity is higher, the influence of the users on the topic is higher. Therefore, the system acquires the hot topic of a fund, the cosine similarity between news under the topic T and the blogs sent by the user, and takes the value as the topic influence w of the user t (i) In that respect If w t (i) The larger the value, the more relevant the user is to the topic. In general, users with higher topic relevance are more likely to be concerned and are more likely to be transferred to the node by other nodes.
In order to solve the problem that each node in the influence strength algorithm cannot set the transition probability in a self-adaptive mode, the invention introduces the calculated influence among the nodes, the user behavior influence and the topic influence, and provides a self-adaptive user influence strength algorithm, namely an AUIR algorithm. The algorithm weights the node transfer probability based on the self-adaptive weight factor, so that certain behaviors, nodes and topic biases can be achieved according to different attributes of specific neighbor nodes during transfer among the nodes. The core formula of the AUIR algorithm is as follows:
Figure BDA0002868343540000081
Figure BDA0002868343540000082
Figure BDA0002868343540000083
equation (6) is the adaptive weight factor calculation equation, where j is the neighbor node of i, w u (j)、w t (j) Respectively the user behavior influence and topic influence, w, of the user j n And (i, j) is the influence between the nodes of the user i and the user j, and m, n and k are the user behavior influence, the topic influence and the weight of the influence between the nodes respectively.
In the formula (7), if the node i is directly connected to the node j, a ij Otherwise, it is 0. The formula is an iterative formula of an AUIR algorithm, and in the user iterative process, a self-adaptive weight factor w (i, j) is added, so that the node transfer process has certain node, behavior and topic bias.
Equation (8) is the AUIR value of Node i in a stable state, after iteration is finished, each Node obtains a stable influence value, and because the influence ranking is not counted in the 'group Node', the traditional method averagely distributes the converged influence value to other nodes. However, in view of differences of each Node, the method improves the formula (2), and distributes the converged influence value of the 'group Node' to each Node according to the weight proportion of each Node.

Claims (5)

1. A fund recommendation method based on social big data is characterized by comprising the following steps:
(1) collecting and processing investment preference information of a user and crawling behavior and public opinion data of a user of a financial social network site;
(2) constructing a dynamic social topological network structure, designing a self-adaptive user influence strength algorithm, and carrying out heat analysis on popular fund products in a financial social network site, wherein the method comprises the following specific steps:
(2.1) adding a background Node in a social user network topological structure, enabling the background Node to be in bidirectional interconnection with all nodes in a graph, and converting an original network G (N, M) with N nodes and M edges into a strongly-connected network with N +1 nodes and M +2N edges;
after introducing the group node, the iterative formula of the node influence is as follows:
Figure FDA0003754949910000011
where i, j is 1,2, …, N +1, and when node i is directly connected to node j, then a ij 1, otherwise is a ij ＝0，
Figure FDA0003754949910000012
Which represents the out-degree of the node j,
Figure FDA0003754949910000013
representing the probability of the node i randomly walking to the node j; all nodes except the background node have an initial value of LR i The background node g is initialized to LR at 1 g ＝0；
The influence value formula of the node i in the steady state is as follows:
Figure FDA0003754949910000014
in the above formula, t c Indicating the number of convergence, LR g (t c ) Obtaining a stable influence value for each node after iteration is finished for an LR value of the node g in a stable state;
(2.2) introducing self-adaptive weight factors into a self-adaptive user influence strength algorithm, wherein the self-adaptive weight factors comprise inter-node influence, user behavior influence and fund heat topic influence;
the calculation formula of the influence between the nodes is as follows:
Figure FDA0003754949910000015
in the above formula, w n (i, j) represents between nodesInfluence with values between 0 and 1, i and j are nodes in the graph, n (i) in And n (j) in The number of nodes pointing to i and j, n (i) out And n (j) out The number of nodes pointed by i and j respectively;
the calculation formula of the user behavior influence is as follows:
Figure FDA0003754949910000021
in the above formula, w u (i) Representing the influence of user behavior, fans (i) representing the number of fans, A i For the user i, the beating frequency in the statistical period, i.e. the activity of the user i in the statistical period, where M i The number of beats of the user i in a T period, T is a statistical period, T i Represents the propagation capability of user i in a specified period, where m j Indicating any blog sent by user i during period T,
Figure FDA0003754949910000022
indicating Bowen m j The forward-to-forward rate of (c),
Figure FDA0003754949910000023
are respectively m j Commenting and like rates of (T) i The method comprises the average praise number, the average comment number and the average forwarding number of a user i in a specified time period T, wherein the higher the activity of the user is, the stronger the post propagation capacity is, the greater the influence of the user is, M is the fan value of the user with the most fans in the users concerning the i, V (i) represents whether the i user is a large V user, if the i user is a large V user, the 1 is adopted, otherwise, the 0 is adopted; a. b and c are respective weights;
the influence of the fund heat topic is obtained by obtaining the heat topic of a fund, the cosine similarity between news under the topic T and the blog sent by the user and taking the value as the influence w of the topic of the user t (i) The self-adaptive weight factor calculation formula is as follows:
Figure FDA0003754949910000024
neighbor nodes where j is i, w u (j)、w t (j) Respectively the user behavior influence and topic influence, w, of the user j n (i, j) is the influence between the nodes of the user i and the user j, and m, n and k are the user behavior influence, the topic influence and the weight of the influence between the nodes respectively;
the iterative formula of the adaptive user influence strength algorithm based on the adaptive weight factor is as follows:
Figure FDA0003754949910000025
node i is directly connected to node j, then a ij 1, otherwise 0; in the user iteration process, adding a self-adaptive weight factor w (i, j) to ensure that the node transfer process has certain node, behavior and topic bias;
under a steady state, the AUIR value formula of the node i based on the adaptive weight factor is as follows:
Figure FDA0003754949910000031
(3) judging the risk of fund investment according to the comparison of the average market profitability value of the current market with historical high-level and low-level numerical values;
(4) and (4) integrating the results of the fund valuation, the market risk analysis and the fund popularity analysis, and screening and recommending fund products meeting the requirements of the clients and friends in the social circle of the clients.
2. The social big data-based fund recommendation method according to claim 1, wherein: and (3) acquiring real-time net value estimation, unit net value and accumulated net value of the fund product, performing incremental crawling operation every day, and judging the risk of fund investment according to the comparison of the average market profitability value of the current market and historical high-level and low-level numerical values.
3. The social big data-based fund recommendation method according to claim 1, wherein: in the step (2), the popularity analysis is combined with the emotion index of the netizen of the financial social network site; the self-adaptive user influence strength algorithm selects the forwarding number, the comment number, the praise number, the vermicelli number and whether the parameters are opinion leader parameters.
4. The social big data-based fund recommendation method according to claim 1, wherein: in the step (4), recommending the investment scheme meeting the requirements of the client to the client, and if the investment scheme does not meet the requirements of the client, repeating the steps (2) and (3) to adjust the combination; the fund product recommended for the customer also supports the purchase of the fund directly on the existing scheme after the friends in the social circle enter the platform in a link sharing mode.
5. The fund recommendation system of the fund recommendation method based on social big data as claimed in claim 1, wherein: the system comprises a financial social contact big data acquisition module, a fund heat degree analysis module, a fund net worth risk analysis module and a fund screening recommendation module;
the financial social big data acquisition module is used for acquiring and processing investment preference information of the user and crawling behavior and public opinion data of the user of the financial social website;
the fund popularity analysis module is used for constructing a dynamic social topological network structure, designing a self-adaptive user influence strength algorithm and carrying out popularity analysis on popular fund products in the financial social network site;
the fund net worth risk analysis module is used for judging the risk size of fund investment according to the comparison of the average market profit value of the current market and historical high-level and low-level numerical values;
and the fund screening and recommending module is used for integrating results of fund evaluation, market risk analysis and fund popularity analysis, and screening and recommending fund products meeting the requirements of customers and friends in social circles of the customers.
CN202011596610.4A 2020-12-29 2020-12-29 Social big data-based fund recommendation system and method Active CN112612942B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011596610.4A CN112612942B (en) 2020-12-29 2020-12-29 Social big data-based fund recommendation system and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011596610.4A CN112612942B (en) 2020-12-29 2020-12-29 Social big data-based fund recommendation system and method

Publications (2)

Publication Number Publication Date
CN112612942A CN112612942A (en) 2021-04-06
CN112612942B true CN112612942B (en) 2022-09-23

Family

ID=75248858

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011596610.4A Active CN112612942B (en) 2020-12-29 2020-12-29 Social big data-based fund recommendation system and method

Country Status (1)

Country Link
CN (1) CN112612942B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112966189B (en) * 2021-04-14 2024-01-26 北京基智科技有限公司 Fund product recommendation system
CN113240472B (en) * 2021-05-19 2024-02-02 深圳索信达数据技术有限公司 Financial product recommendation method, electronic equipment and storage medium
CN114154075B (en) * 2022-02-08 2022-05-17 北京大氪信息科技有限公司 Hot information determination method, device, computer equipment and medium
CN115131155B (en) * 2022-05-11 2023-12-19 李益非 Cloud platform based on investment management

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7373324B1 (en) * 1999-10-07 2008-05-13 Robert C. Osborne Method and system for exchange of financial investment advice
CN105117422A (en) * 2015-07-30 2015-12-02 中国传媒大学 Intelligent social network recommender system
CN107679239A (en) * 2017-10-27 2018-02-09 天津理工大学 Recommend method in a kind of personalized community based on user behavior

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200357071A1 (en) * 2014-09-05 2020-11-12 Vested Interest Co. System and method for personal investing

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7373324B1 (en) * 1999-10-07 2008-05-13 Robert C. Osborne Method and system for exchange of financial investment advice
CN105117422A (en) * 2015-07-30 2015-12-02 中国传媒大学 Intelligent social network recommender system
CN107679239A (en) * 2017-10-27 2018-02-09 天津理工大学 Recommend method in a kind of personalized community based on user behavior

Also Published As

Publication number Publication date
CN112612942A (en) 2021-04-06

Similar Documents

Publication Publication Date Title
CN112612942B (en) Social big data-based fund recommendation system and method
US11659050B2 (en) Discovering signature of electronic social networks
Courtois et al. Farmer bargaining power and market information services
CN111125453B (en) Opinion leader role identification method in social network based on subgraph isomorphism and storage medium
Yang et al. Friend or frenemy? Predicting signed ties in social networks
CN103377250B (en) Top k based on neighborhood recommend method
CN105117422A (en) Intelligent social network recommender system
US20120041769A1 (en) Requests for proposals management systems and methods
CN104254851A (en) Method and system for recommending content to a user
CN104254852A (en) Method and system for hybrid information query
Xu et al. A novel approach to online social influence maximization
CN111143704B (en) Online community friend recommendation method and system integrating user influence relationship
CN111695084A (en) Model generation method, credit score generation method, device, equipment and storage medium
Wu et al. Optimization matrix factorization recommendation algorithm based on rating centrality
Li SWOT analysis of e‐commerce development of rural tourism farmers' professional cooperatives in the era of big data
Deng et al. Recommender system for marketing optimization
CN115310722A (en) Agricultural product price prediction method based on data statistics
Juszczuk et al. Real-world data difficulty estimation with the use of entropy
Liu et al. REDRL: A review-enhanced Deep Reinforcement Learning model for interactive recommendation
US20140280216A1 (en) Automated ranking of contributors to a knowledge base
León et al. Tackling travel behaviour: an approach based on fuzzy cognitive maps
Karthikeyan et al. Machine learning techniques application: social media, agriculture, and scheduling in distributed systems
GB2517358A (en) Recommendation creation system
Zhao et al. Predictive models with end user preference
Li et al. Case-Based Reasoning for Personalized Recommender on User Preference through Dynamic Clustering

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
GR01 Patent grant
GR01 Patent grant