CN113570455A - Stock recommendation method and device, computer equipment and storage medium - Google Patents
Stock recommendation method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN113570455A CN113570455A CN202010357317.6A CN202010357317A CN113570455A CN 113570455 A CN113570455 A CN 113570455A CN 202010357317 A CN202010357317 A CN 202010357317A CN 113570455 A CN113570455 A CN 113570455A
- Authority
- CN
- China
- Prior art keywords
- stock
- target
- historical
- model
- group
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000003860 storage Methods 0.000 title claims abstract description 16
- 239000011159 matrix material Substances 0.000 claims abstract description 84
- 238000012549 training Methods 0.000 claims abstract description 63
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 16
- 238000011156 evaluation Methods 0.000 claims description 13
- 238000013139 quantization Methods 0.000 claims description 8
- 238000011946 reduction process Methods 0.000 claims description 6
- 230000003442 weekly effect Effects 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 description 22
- 238000010801 machine learning Methods 0.000 description 18
- 230000000694 effects Effects 0.000 description 11
- 230000002411 adverse Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 238000009826 distribution Methods 0.000 description 8
- 238000012163 sequencing technique Methods 0.000 description 5
- 238000004590 computer program Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
Images
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
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
-
- 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
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Finance (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Accounting & Taxation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Algebra (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a stock recommendation method, which comprises the following steps: carrying out stock portrait feature extraction on the historical stock information of the target stock group to obtain a historical portrait of each stock; obtaining a stock selection target characteristic matrix of a target stock group according to the historical portrait and the historical earning rate of each stock; pre-training a pre-stored deep text matching model according to the historical portrait of each stock and the stock selection target characteristic matrix, and selecting a target model; training a target model according to the historical picture of each stock and the stock selection target characteristic matrix; and obtaining the historical stock information of the forecast time and the stock group to be recommended, inputting the historical stock information into the target model, and obtaining the sorting result of the stock group to be recommended at the forecast time based on the stock selecting target characteristic. The method can show high application value and robustness in the ranking application based on the stock profitability prediction. In addition, the invention also provides a stock recommendation device, a computer device and a computer readable storage medium.
Description
Technical Field
The invention relates to the technical field of financial investment, in particular to a stock recommendation method and device, computer equipment and a storage medium.
Background
Stock is a common investment mode, and users can obtain income when market valuation fluctuates upwards by investing principal funds, and the dream of each stockholder is that the stock obtains as much income as possible.
But stock markets are ever-changing, and no scheme with stability and high application value is provided for predicting the future trend of stocks at present. Accurate prediction is difficult to achieve by both stock evaluators and manual analysis, and the unknown and risk of the stock market are still a difficult direction to overcome.
Therefore, a method and an apparatus for recommending stocks, a computer device, and a storage medium are needed.
Disclosure of Invention
Technical problem to be solved
In view of the problems in the art described above, the present invention is at least partially addressed. Therefore, the invention aims to provide a stock recommendation method which can show high application value and robustness in the ranking application based on stock earnings.
The second purpose of the invention is to provide a stock recommendation device.
A third object of the invention is to propose a computer device.
A fourth object of the invention is to propose a computer-readable storage medium.
(II) technical scheme
In order to achieve the above object, an aspect of the present invention provides a stock recommendation method, including the following steps:
acquiring historical stock information of a target stock group;
extracting the characteristics of the stock portrait according to the historical stock information to obtain the historical portrait of each stock in the target stock group;
calculating the corresponding stock selection target characteristic of each stock according to the historical portrait of each stock and the historical earning rate of each stock so as to obtain a stock selection target characteristic matrix of a target stock group;
pre-training each pre-stored deep text matching model according to the historical portrait of each stock and the stock selection target characteristic matrix to select a target model;
training a target model according to the historical picture of each stock and the stock selection target characteristic matrix;
and acquiring the forecast time and the historical stock information of the stock group to be recommended, and inputting the forecast time and the historical stock information of the stock group to be recommended into the target model to obtain a sorting result of the stock group to be recommended at the forecast time based on the stock selecting target characteristic.
The stock recommendation method provided by the embodiment of the invention firstly obtains the historical stock information of the target stock group, and extracting the stock portrait characteristics according to the historical stock information to obtain the historical portrait of each stock in the target stock group, then, the corresponding stock selection target characteristic of each stock is calculated according to the historical portrait of each stock and the historical earning rate of each stock ticket so as to obtain a stock selection target characteristic matrix of the target stock group, then pre-training each pre-stored deep text matching model according to the historical portrait of each stock and the stock selection target matrix to select a target model, finally training the target model according to the historical portrait of each stock and the stock selection target characteristic matrix, therefore, when the historical stock information of the forecast time and the stock group to be recommended is input into the target model, the sorting result of the stock group to be recommended at the forecast time based on the stock selecting target characteristic can be obtained. Therefore, the invention converts the stock selecting problem in the quantitative investment into the matching problem of the target stock group and the stock selecting target in the machine learning, constructs the stock selecting target according to the stock history picture and the stock history profitability, represents the stock selecting target as the problem in the machine learning question-answering problem, creates the condition for realizing the problem definition to the problem solution, provides the support for training the high-quality target model, pre-trains each pre-stored deep text matching model according to the history picture of each stock and the stock selecting target matrix to select the target model suitable for the current stock selecting question-answering so as to reduce the adverse effect of the non-independent and same-distribution characteristic of the training data, thereby inputting the forecasting time and the history stock information of the stock group to be recommended into the trained target model, and accurately forecasting the sequencing result based on the stock profitability, the accuracy of the algorithm is greatly improved, and the method has high application value and robustness.
Optionally, the historical stock information includes stock market trading data and financial social media data, wherein the stock sketch feature extraction according to the historical stock information includes: and extracting traditional quantization factors from stock market trading data and extracting novel social media quantization factors from financial social media data from the perspective of asset pricing.
Optionally, calculating the stock-selecting target feature of each stock according to the historical representation of each stock and the historical profitability of each stock, including:
get each stock
Historical portrait feature matrix
And each stock
Historical rate of return set
The adopted time window is the degree of week which comprises all transaction days in the calendar week, the historical portrait characteristic matrix is obtained by backtracking w windows on the time sequence by the stock portrait characteristic vector, w belongs to N + and tiAt a certain week time point, j is the jth stock in the target group of stocks, j<n, n is the number of stocks in the target group stock ticket,
the characteristic vector of the weekly image of the stock is taken as the characteristic vector of the weekly image of the stock;
descending order of R(j)And from
Selecting the stock image feature vector corresponding to k earnings rate before the rank to obtain each stock
The stock-selecting target feature matrix
Wherein,
is a k × m dimensional matrix; k is the time window of optimum yield, k is a,1<a≤w。
Optionally, obtaining a stock selection target feature matrix of the target stock group includes:
adopting t-SNE model to select stock target characteristic matrix of each stock
Performing a first dimensionality reduction process to obtain a stock selection intermediate matrix of the target stock set
Using PCA model to select intermediate matrix
Performing a second dimensionality reduction process to obtain a stock selection target feature matrix of the target stock group
Wherein, the first dimension reduction parameter dfB ∈ N +; second dimensionality reduction parameter ds＝ if(m,c),m<c<(b×n)。
Optionally, pre-training each pre-stored deep text matching model according to the historical representation of each stock and the stock selection target feature matrix to select a target model, including:
according to the historical portrait of each stock and the stock-selecting target characteristic matrix, pre-training each pre-stored deep text matching model by combining all hyper-parameters of the model;
in the pre-training process, evaluating each hyper-parameter combination according to the model performance to obtain a hyper-parameter combination evaluation value set of each model;
calculating a first standard deviation of each model according to the hyper-parameter combination evaluation value set of each model to obtain a first standard deviation set of all models;
and calculating a second standard deviation according to the first standard deviation set, and selecting a target model according to the second standard deviation.
Optionally, selecting the target model according to the second standard deviation includes:
judging the second standard deviation;
if the second standard deviation is smaller than a preset first threshold, taking a model corresponding to the minimum standard deviation in the first standard deviation set as a target model;
if the second standard deviation is greater than or equal to a preset first threshold and less than or equal to a preset second threshold, taking a model corresponding to the maximum standard deviation in the first standard deviation set as a target model;
and if the second standard deviation is larger than a preset second threshold value, taking a model corresponding to the middle standard deviation in the first standard deviation set as a target model.
Optionally, training the target model according to the historical sketch of each stock and the stock-selecting target feature matrix comprises:
selecting the hyper-parameter set with the largest evaluation value from the hyper-parameter combination evaluation value set of the target model to cooperate as the optimal hyper-parameter combination of the target model;
and training the target model according to the optimal hyper-parameter combination.
In order to achieve the above object, another aspect of the present invention provides a stock recommendation apparatus, including:
the acquisition module is used for acquiring historical stock information of the target stock group;
the stock portrait module is used for extracting the stock portrait characteristics according to the historical stock information so as to obtain the historical portrait of each stock in the target stock group;
the stock selecting characteristic extracting module is used for calculating the corresponding stock selecting target characteristic of each stock according to the historical picture of each stock and the historical yield of each stock so as to obtain a stock selecting target characteristic matrix of the target group stock ticket;
the model selection module is used for pre-training each pre-stored depth text matching model according to the historical portrait of each stock and the stock selection target characteristic matrix so as to select a target model;
the model training module is used for training a target model according to the historical portrait of each stock and the stock selection target characteristic matrix;
the acquisition module is also used for acquiring the forecast time and the historical stock information of the stock group to be recommended, and inputting the forecast time and the historical stock information of the stock group to be recommended into the target model so as to obtain a sorting result of the stock group to be recommended based on the stock-selecting target characteristic in the forecast time.
The stock recommending device provided by the embodiment of the invention obtains the historical stock information of a target stock group through an obtaining module, obtains the historical stock characteristic of each stock in the target stock group by performing stock sketch characteristic extraction according to the historical stock information through a stock sketch module, calculates the corresponding stock selection target characteristic of each stock according to the historical sketch of each stock and the historical profitability of each stock through a stock selection characteristic extracting module to obtain a stock selection target characteristic matrix of the target stock group, pre-trains each pre-stored depth text matching model according to the historical sketch of each stock and the stock selection target matrix through a model selecting module to select a target model, and trains the target model according to the historical sketch of each stock and the stock selection target characteristic matrix through a model training module, thus, when the historical stock information of the forecast time and the stock group to be recommended is input into the target model, the ranking result of the stock group to be recommended at the forecast time based on the stock-selecting target characteristic can be obtained. It can be seen that the invention converts the stock-selecting problem in the quantitative investment into the matching problem of the target stock group and the stock-selecting target in the machine learning, constructs the stock-selecting target according to the stock history picture and the stock history profitability, represents the stock-selecting target as the problem in the machine learning question-answer problem, creates the condition for realizing the problem definition to the problem solution, provides the support for training the high-quality target model, and pre-trains each pre-stored deep text matching model according to the history picture and the stock-selecting target matrix of each stock to select the target model suitable for the current stock-selecting question-answer, so as to reduce the adverse effect of the non-independent and same-distribution characteristic of the training data, thereby the invention inputs the history stock information of the predicted time and the stock group to be recommended into the trained target model, can accurately predict the ranking result based on the stock profitability, the accuracy of the algorithm is greatly improved, and high application value and robustness are shown.
In addition, the embodiment of the invention also provides a computer device which comprises a memory, a processor and a stock recommendation program which is stored on the memory and can be run on the processor, and when the processor executes the stock recommendation program, the stock recommendation method is realized.
The computer equipment of the embodiment of the invention can input the forecast time and the historical stock information of the stock group to be recommended into a trained target model by running the stock recommendation program stored in the memory through the processor, accurately forecast the sequencing result based on the profitability of the stocks, greatly improve the accuracy of the algorithm and show high application value and robustness, because the invention converts the stock selection problem in the quantitative investment into the matching problem of the target stock group and the stock selection target in the machine learning, constructs the stock selection target according to the historical stock figure and the historical profitability of the stocks, takes the stock selection target as the problem representation in the question of the machine learning question, creates conditions for realizing the problem definition to the problem solution, also provides support for training a high-quality target model, and pre-trains each pre-stored deep text matching model according to the historical figure of each stock and the stock selection target matrix, the target model suitable for the current stock-selecting question-answer is selected so as to reduce the adverse effect of the non-independent and same-distribution characteristics of the training data.
Furthermore, an embodiment of the present invention also provides a computer-readable storage medium, on which a stock recommendation program is stored, which, when executed by a processor, implements the stock recommendation method as described above.
The computer readable storage medium provided by the embodiment of the invention can input the prediction time and the historical stock information of the stock group to be recommended into a trained target model when a stock recommendation program stored on the computer readable storage medium is executed by a processor, accurately predict the ranking result based on the stock profitability, greatly improve the accuracy of the algorithm and show high application value and robustness And selecting a target model suitable for the current stock-selecting question and answer so as to reduce the adverse effect of the non-independent and same-distribution characteristics of the training data.
(III) advantageous effects
The invention has the beneficial effects that:
the stock recommendation method and device provided by the invention convert the stock selection problem in the quantitative investment into the matching problem of the target stock group and the stock selection target in the machine learning, construct the stock selection target according to the stock history picture and the stock history profitability, represent the stock selection target as the problem in the machine learning question-answer problem, create conditions for realizing the problem definition to the problem solution, also provide support for training a high-quality target model, and pre-train each pre-stored deep text matching model according to the history picture of each stock and the stock selection target stock matrix to select a target model suitable for the current stock selection question-answer, so as to reduce the adverse effect of the non-independent and same-distribution characteristics of the training data, thereby the invention inputs the forecast time and the history stock information of the stock group to be recommended into the trained target model, the method can accurately predict the ranking result based on the stock profitability, greatly improves the accuracy of the algorithm, and shows high application value and robustness.
Drawings
The invention is described with the aid of the following figures:
FIG. 1 is a flow diagram of a stock recommendation method according to one embodiment of the invention;
FIG. 2 is a design framework diagram of the MV-LSTM model according to one embodiment of the invention;
fig. 3 is a block diagram illustrating a stock recommendation apparatus according to an embodiment of the present invention.
[ description of reference ]
1: an acquisition module;
2: a stock sketch module;
3: a stock selection feature extraction module;
4: a model selection module;
5: and a model training module.
Detailed Description
For the purpose of better explaining the present invention and to facilitate understanding, the present invention will be described in detail by way of specific embodiments with reference to the accompanying drawings.
The stock recommendation method and the device provided by the embodiment of the invention convert the stock selection problem in quantitative investment into the matching problem of the target stock group and the stock selection target in machine learning, provide a stock selection target algorithm and a deep text matching model selection algorithm, can greatly improve the accuracy of the algorithm in the sequencing application based on stock earning rate prediction, and show high application value and robustness.
In order to better understand the above technical solutions, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
A stock recommendation method and a stock recommendation apparatus according to an embodiment of the present invention will be described with reference to the accompanying drawings.
Fig. 1 is a flowchart illustrating a stock recommendation method according to an embodiment of the present invention.
As shown in fig. 1, the stock recommendation method includes the steps of:
Wherein the historical stock information includes stock market trading data and financial social media data.
Specifically, as one example, a set of stocks may be screened from Shanghai deep stock market in China as a target set of stocks. The data sources of the historical stock Information of the target stock group may include a Wind Financial Terminal (WFT), a data of contacts (DATAYES), and an East Money Information network (East Money Information). The Wind financial terminal is one of the most qualified financial information service providers in China, and provides accurate, timely and complete Shanghai-Shen stock data, including basic information, quotation, equity, company management, transaction data, major events, financial data, index data and the like; the general data is one of the most professional financial big data providers in China, and provides a quantitative factor library with comprehensive types and reliable quality; the oriental wealth network is one of the finance and finance security portal websites with the largest Chinese access amount and the largest influence, and is a stock bar (stock topic community) under the flag to provide real-time quotation comments and individual stock exchange forums for users. Of course, the data source of the historical stock information of the target stock group may also include data such as boudouard or google search trends or indices.
And step 102, preprocessing historical stock information.
The preprocessing step comprises data cleaning, denoising, integration and the like.
Specifically, as one embodiment, stock portrait feature extraction is performed according to historical stock information, and comprises the following steps: from an asset pricing perspective, traditional quantization factors are extracted from stock market trading data, and new social media quantization factors are extracted from financial social media data (as shown in table 1). Factors influencing the stock income are systematically and comprehensively quantified from the perspective of asset pricing, so that the financial attributes of the risky asset are comprehensively described, and a data base is provided for solving subsequent problems.
TABLE 1 stock sketch
And step 104, calculating the corresponding stock selection target characteristic of each stock according to the historical portrait of each stock and the historical earning rate of each stock so as to obtain a stock selection target characteristic matrix of the target stock group.
Specifically, as an embodiment, calculating the stock-selecting target characteristic of each stock according to the historical portrait of each stock and the historical earning rate of each stock includes:
get each stock
Historical portrait feature matrix
And each stock
Historical rate of return set
The adopted time window is the degree of week which comprises all transaction days in the calendar week, the historical portrait characteristic matrix is obtained by backtracking w windows on the time sequence by the stock portrait characteristic vector, w belongs to N + and tiAt a certain week time point, j is the jth stock in the target group of stocks, j<n, n is the number of stocks in the target group stock ticket,
the characteristic vector of the weekly image of the stock is taken as the characteristic vector of the weekly image of the stock; descending order of R(j)And from
Selecting stock sketch feature vectors corresponding to k earnings before ranking to obtain each stock
The stock-selecting target feature matrix
Wherein,
is a k × m dimensional matrix; k is the time window of optimum yield, k is a,1<a≤w。
Further, as an embodiment, obtaining a stock selection target feature matrix of a target stock group includes: adopting t-SNE model to select stock target characteristic matrix of each stock
Performing a first dimensionality reduction process to obtain a stock selection intermediate matrix of the target stock group
Selecting a middle matrix by PCA model
Performing a second dimensionality reduction to obtain a stock selection target feature matrix of the target stock group
Wherein, the first dimension reduction parameter dfB ∈ N +; second dimensionality reduction parameter ds＝ if(m,c),m<c<(b × n). The dimensionality reduction processing is carried out on the stock selection target characteristic matrix of each stock, so that the operation amount can be greatly reduced, and the processing speed is improved.
And 105, pre-training each pre-stored deep text matching model according to the historical picture of each stock and the stock selection target characteristic matrix to select a target model.
Specifically, as one embodiment, step 105 includes: according to the historical picture of each stock and the stock-selecting target characteristic matrix, pre-training each pre-stored deep text matching model by all hyper-parameter combinations thereof; in the pre-training process, each hyper-parameter combination is evaluated according to the model performance to obtain a hyper-parameter combination evaluation value set k of each model i(i){ Score1, Score2, …, Scorek }, k being the number of hyper-parametric combinations; evaluating a set of values A from a hyper-parametric combination of each model(i)Calculating the first standard deviation of each model
To obtain a first set of standard deviations for all models
Calculating a second standard deviation sigma from the first set of standard deviationsBAnd according to the second standard deviation sigmaBAnd selecting the target model. Wherein, in the model training process, the stock-selecting target characteristic matrix is regarded as a machineThe problem representation in question-answering questions is learned by the machine, and the historical representation of each stock is regarded as the answer representation in question-answering questions of machine learning.
Further, as an embodiment, according to the second standard deviation σBSelecting a target model, comprising: standard deviation of two pairsBJudging; if the second standard deviation sigmaBIf the standard deviation is smaller than a preset first threshold value a, taking a model corresponding to the minimum standard deviation Min (B) in the first standard deviation set B as a target model; if the second standard deviation sigmaBIf the standard deviation is greater than or equal to a preset first threshold value a and less than or equal to a preset second threshold value B, taking a model corresponding to the maximum standard deviation Max (B) in the first standard deviation set B as a target model; if the second standard deviation sigmaBAnd if the standard deviation is larger than a preset second threshold B, taking a model corresponding to the intermediate standard deviation Median (B) in the first standard deviation set B as a target model.
Further, as an embodiment, the preset first threshold a is 0.01, and the preset second threshold b is 0.035.
The target model selection algorithm provided by the invention can select a target model suitable for the current stock selection question and answer according to the characteristics of text matching models with different depths so as to reduce the adverse effect of the non-independent and same distribution characteristics of training data, thereby inputting the prediction time and the historical stock information of the stock group to be recommended into the trained target model, accurately predicting the ranking result based on the stock profitability, greatly improving the accuracy of the algorithm and showing high application value and robustness.
Specifically, as one example, the pre-stored deep text matching models may include a DRMM-TKS model, a CONV-KNRM model, and an MV-LSTM model. For the target set of stocks screened from Shanghai-deep stock market in China, MV-LSTM is better at capturing signals in time series data and performs best in model pre-training.
For the MV-LSTM model, the details are as follows:
as shown in FIG. 2, the MV-LSTM model first generates a multi-position sentence representation using a Bidirectional short-length memory algorithm (Bi-LSTMs). Bi-LSTMs can beCapturing information in the backward and forward directions of the sequence, the output at each position being two vectors from the two directions
And
the connection of (2). If a sentence T ═ T is defined(1),t(2),…,t(k),…,t(w)As algorithm input, t(k)Representing the word vector representation at position t. After the calculation of the formulas (1-1), (1-2), (1-3), (1-4) and (1-5), the position t outputs a word vector expression h(t)。
In the formula, i, f and o respectively represent an input layer gate, a forgetting gate and an output layer gate; c is a memory unit for storing information.
Then, the sentence at the t-th position is represented by
And
connection generation
Therein, (.)TIndicating the permutation operation to be used in the next operation. Then, on the basis of position sentence representation, modeling is carried out on interactive sentences from different positions, and MV-LSTM adopts three similarity functions, namely cosine similarity, bilinear similarity and tensor layer similarity, and carries out pair matching
And
is modeled, wherein
And
respectively represent sentences SXAnd SYThe position sentences at i and j indicate. And finally, extracting the first k strongest local information from the similarity matrix or the similarity tensor by the k-Max pooling layer according to a formula (1-6), generating a higher-level vector representation r through a full-connection hidden layer, and calculating a matching score by the multi-layer perceptron according to formulas (1-7) and (1-8).
s＝Wsr+bs, (1-8)
The formula for calculating the loss function of MV-LSTM is:
in the formula,
representing positive and negative examples, respectively.
And 106, training a target model according to the historical picture of each stock and the stock selection target characteristic matrix.
Specifically, as one embodiment, step 106 includes: evaluating a set of values A from a hyper-parametric combination of target models(i)Selecting the hyper-parameter combination with the largest evaluation value as the optimal hyper-parameter combination of the target model; and training the target model according to the optimal hyper-parameter combination.
And step 107, obtaining the historical stock information of the forecast time and the stock group to be recommended, and inputting the historical stock information of the forecast time and the stock group to be recommended into the target model to obtain a sorting result of the stock group to be recommended at the forecast time based on the stock-selecting target characteristic.
Specifically, as an example, for the target stock group screened from Shanghai-Shen stock market in China, the invention trains the MV-LSTM model, and in the process of predicting the stock profitability ranking of the stock group to be recommended on a certain turning date, the trained MV-LSTM model obtains good performance under two evaluation indexes of NDCG @10 and ERR.
In conclusion, the stock recommendation method provided by the invention converts the stock selection problem in quantitative investment into the matching problem of the target stock group and the stock selection target in machine learning, and combines the provided stock selection target algorithm and the deep text matching model selection algorithm, so that the accuracy of the algorithm can be greatly improved in the ranking application based on the stock profit rate prediction, and the high application value and the high stability are shown.
According to the stock recommendation method provided by the embodiment of the invention, firstly, the history stock information of the target stock group is obtained, and extracting the stock image characteristics according to the historical stock information to obtain the historical image of each stock in the target stock group, then, the corresponding stock selection target characteristic of each stock is calculated according to the historical portrait of each stock and the historical earning rate of each stock to obtain a stock selection target characteristic matrix of the target stock group, then pre-training each pre-stored deep text matching model according to the historical portrait of each stock and the stock selection target matrix to select a target model, finally training the target model according to the historical portrait of each stock and the stock selection target characteristic matrix, therefore, when the historical stock information of the forecast time and the stock group to be recommended is input into the target model, the sorting result of the stock group to be recommended at the forecast time based on the stock selecting target characteristic can be obtained. Therefore, the invention converts the stock selecting problem in the quantitative investment into the matching problem of the target stock group and the stock selecting target in the machine learning, constructs the stock selecting target according to the stock history picture and the stock history profitability, represents the stock selecting target as the problem in the machine learning question-answering problem, creates the condition for realizing the problem definition to the problem solution, provides the support for training the high-quality target model, pre-trains each pre-stored deep text matching model according to the history picture of each stock and the stock selecting target matrix to select the target model suitable for the current stock selecting question-answering, and reduces the adverse effect of the non-independent homodistributional characteristic of the training data, thereby inputting the prediction time and the history stock information of the stock group to be recommended into the trained target model, and accurately predicting the sequencing result based on the stock profit rate, the accuracy of the algorithm is greatly improved, and the method has high application value and robustness.
Fig. 3 is a block diagram of a stock recommendation device according to an embodiment of the present invention.
As shown in fig. 3, the stock recommendation apparatus includes: the stock sketch model comprises an acquisition module 1, a stock sketch module 2, a stock selection feature extraction module 3, a model selection module 4 and a model training module 5.
The acquisition module 1 is used for acquiring historical stock information of a target stock group; the stock portrait module 2 is used for extracting the stock portrait characteristics according to the historical stock information so as to obtain the historical portrait of each stock in the target stock group; the stock selecting characteristic extracting module 3 is used for calculating the corresponding stock selecting target characteristic of each stock according to the historical portrait of each stock and the historical earning rate of each stock so as to obtain a stock selecting target characteristic matrix of the target group of stocks; the model selection module 4 is used for pre-training each pre-stored depth text matching model according to the historical portrait of each stock and the stock selection target characteristic matrix so as to select a target model; the model training module 5 is used for training a target model according to the historical portrait of each stock and the stock selection target characteristic matrix; the obtaining module 1 is further configured to obtain the prediction time and the historical stock information of the stock group to be recommended, and input the prediction time and the historical stock information of the stock group to be recommended into the target model, so as to obtain a sorting result of the stock group to be recommended at the prediction time based on the stock-selecting target feature.
As one example, the stock sketch module 2 is specifically configured to extract traditional quantization factors from stock market trading data and new social media quantization factors from financial social media data from an asset pricing perspective to obtain a historical sketch of each stock in the target stock group.
As an embodiment, the stock-selecting feature extraction module 3 is specifically configured to arrange each stock ticket in descending order
Historical yield set R of(j)And from each stock
Historical portrait feature matrix
Selecting stock portrait eigenvectors corresponding to j profitability before ranking to obtain each stock ticket
The stock-selecting target feature matrix
Then adopting t-SNE model to select stock target characteristic matrix of each stock
Performing a first dimensionality reduction process to obtain a stock selection intermediate matrix of the target stock set
Using PCA model to select intermediate matrix
Performing a second dimensionality reduction to obtain a stock selection target feature matrix of the target stock group
As an embodiment, the model selection module 4 is specifically configured to pre-train each pre-stored deep text matching model with all hyper-parameter combinations thereof according to the historical image of each stock and the stock selection target feature matrix; in the pre-training process, each hyper-parameter combination is evaluated according to the model performance to obtain a hyper-parameter combination evaluation value set A of each model i(i){ Score1, Score2, …, Scorek }, k being the number of hyper-parametric combinations; evaluating a set of values A from a hyper-parametric combination of each model(i)Calculating the first standard deviation of each model
To obtain a first set of standard deviations for all models
Calculating a second standard deviation sigma from the first set of standard deviationsBAnd according to the second standard deviation sigmaBAnd selecting the target model.
As an example, according to the second standard deviation σ, the model selection module 4BSelecting a target model, comprising: standard deviation of two pairsBJudging; if the second standard deviation sigmaBIf the standard deviation is smaller than a preset first threshold value a, taking a model corresponding to the minimum standard deviation Min (B) in the first standard deviation set B as a target model; if the second standard deviation sigmaBIf the standard deviation is greater than or equal to a preset first threshold value a and less than or equal to a preset second threshold value B, taking a model corresponding to the maximum standard deviation Max (B) in the first standard deviation set B as a target model; if the second standard deviation sigmaBAnd if the standard deviation is larger than a preset second threshold B, taking a model corresponding to the intermediate standard deviation Median (B) in the first standard deviation set B as a target model.
As an embodiment, the model training module 5 is specifically configured to evaluate the set of values A from a hyper-parameter combination of the target model(i)Selecting the hyper-parameter combination with the largest evaluation value as the optimal hyper-parameter combination of the target model; combining training objectives according to optimal hyper-parametersAnd (5) marking the model.
In conclusion, the stock recommendation device provided by the invention converts the stock selection problem in quantitative investment into the matching problem of the target stock group and the stock selection target in machine learning, and combines the provided stock selection target algorithm and the deep text matching model selection algorithm, so that the accuracy of the algorithm can be greatly improved in the sequencing application based on stock profit rate prediction, and the high application value and the high stability are shown.
According to the stock recommendation device provided by the embodiment of the invention, historical stock information of a target stock set is obtained through an obtaining module, a stock image module is used for extracting stock image characteristics according to the historical stock information to obtain a historical image of each stock in the target stock set, then a stock selection characteristic extracting module is used for calculating the stock selection target characteristics of each corresponding stock according to the historical image of each stock and the historical profitability of each stock to obtain a stock selection target characteristic matrix of the target stock set, then a model selecting module is used for pre-training each pre-stored deep text matching model according to the historical image of each stock and the stock selection target matrix to select a target model, and finally a model training module is used for training the target model according to the historical image of each stock and the stock selection target characteristic matrix, therefore, when the historical stock information of the forecast time and the stock group to be recommended is input into the target model, the sorting result of the stock group to be recommended at the forecast time based on the stock selecting target characteristic can be obtained. It can be seen that the invention converts the stock-selecting problem in the quantitative investment into the matching problem of the target stock group and the stock-selecting target in the machine learning, constructs the stock-selecting target according to the historical stock figure and the historical profit rate of the stocks, represents the stock-selecting target as the problem in the machine learning question-answer problem, creates the condition for realizing the problem definition to the problem solution, provides the support for training the high-quality target model, and pre-trains each pre-stored deep text matching model according to the historical figure of each stock and the stock-selecting target matrix to select the target model suitable for the current stock-selecting question-answer, so as to reduce the adverse effect of the non-independent same distribution characteristic of the training data, thereby the invention inputs the pre-measured time and the historical stock information of the stock group to be recommended into the trained target model, can accurately predict the ranking result based on the stock profit rate, the accuracy of the algorithm is greatly improved, and the method has high application value and robustness.
In addition, the embodiment of the invention also provides a computer device which comprises a memory, a processor and a stock recommendation program which is stored on the memory and can be run on the processor, and when the processor executes the stock recommendation program, the stock recommendation method is realized.
According to the computer equipment provided by the embodiment of the invention, the stock recommendation program stored in the memory is run by the processor, the forecast time and the historical stock information of the stock group to be recommended can be input into the trained target model, the ranking result based on the stock profitability is accurately forecasted, the accuracy of the algorithm is greatly improved, the high application value and the robustness are shown, because the invention converts the stock selection problem in the quantitative investment into the matching problem of the target stock group and the stock selection target in the machine learning, constructs the stock selection target according to the stock history picture and the stock historical profitability, takes the stock selection target as the problem representation in the machine learning question-answering problem, creates conditions for realizing the problem definition to the problem solution, also provides support for training the high-quality target model, and pre-trains each pre-stored deep text matching model according to the historical picture and the stock selection target matrix of each stock, the target model suitable for the current stock-selecting question-answer is selected so as to reduce the adverse effect of the non-independent and same-distribution characteristics of the training data.
Furthermore, an embodiment of the present invention also provides a computer-readable storage medium, on which a stock recommendation program is stored, which, when executed by a processor, implements the stock recommendation method as described above.
According to the computer readable storage medium provided by the embodiment of the invention, when a stock recommendation program stored on the computer readable storage medium is executed by a processor, the stock recommendation program can input the prediction time and the historical stock information of a stock group to be recommended into a trained target model, accurately predict the ranking result based on the stock profitability, greatly improve the accuracy of the algorithm, and show high application value and robustness And training to select a target model suitable for the current stock selection question and answer so as to reduce the adverse effect of the non-independent and identically distributed characteristics of the training data.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions.
It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the terms first, second, third and the like are for convenience only and do not denote any order. These words are to be understood as part of the name of the component.
Furthermore, it should be noted that in the description of the present specification, the description of the term "one embodiment", "some embodiments", "examples", "specific examples" or "some examples", etc., means that a specific feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, the claims should be construed to include preferred embodiments and all changes and modifications that fall within the scope of the invention.
It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the invention. Thus, the present invention should also include such modifications and variations provided they come within the scope of the claims and their equivalents.
Claims (10)
1. A stock recommendation method, comprising the steps of:
acquiring historical stock information of a target stock group;
extracting the characteristics of the stock portrait according to the historical stock information to obtain the historical portrait of each stock in the target stock group;
calculating the corresponding stock selection target characteristic of each stock according to the historical portrait of each stock and the historical earning rate of each stock so as to obtain a stock selection target characteristic matrix of the target stock group;
pre-training each pre-stored deep text matching model according to the historical portrait of each stock and the stock selection target characteristic matrix to select a target model;
training the target model according to the historical portrait of each stock and the stock selection target characteristic matrix;
and acquiring the forecast time and the historical stock information of the stock group to be recommended, and inputting the forecast time and the historical stock information of the stock group to be recommended into the target model to obtain a sorting result of the stock group to be recommended at the forecast time based on the stock-selecting target characteristic.
2. The stock recommendation method of claim 1, wherein the historical stock information comprises stock market trading data and financial social media data, and wherein performing stock portrait feature extraction based on the historical stock information comprises:
and extracting traditional quantization factors from the stock market trading data and extracting novel social media quantization factors from the financial social media data from the view point of asset pricing.
3. The stock recommendation method of claim 1, wherein calculating the stock-selection target feature of each stock according to the historical representation of each stock and the historical profitability of each stock comprises:
get each stock
Historical portrait feature matrix
And each stock
Historical rate of return set
The time window is used as the degree of week, the degree of week comprises all the trading days in the calendar week, the historical portrait characteristic matrix is obtained by backtracking w windows on the time sequence by the stock portrait characteristic vector, w belongs to N + and tiAt a certain week time point, j is the jth stock in the target group of stocks, j<n, n is the number of stocks in the target group of stocks,
the characteristic vector of the weekly image of the stock is taken as the characteristic vector of the weekly image of the stock;
descending order of R(j)And from
Selecting stock sketch feature vectors corresponding to k earnings before ranking to obtain each stock
The stock-selecting target feature matrix
Wherein,
is a k × m dimensional matrix; k is the time window of optimum yield, k is a,1<a≤w。
4. The stock recommendation method of claim 3, wherein obtaining a target feature matrix for the selection of stocks of the target group of stocks comprises:
adopting t-SNE model to select stock target characteristic matrix of each stock
Performing a first dimensionality reduction process to obtain a stock selection intermediate matrix of the target stock set
Applying PCA model to the selected intermediate matrix
Performing a second dimensionality reduction process to obtain a stock selection target feature matrix of the target stock group
Wherein, the first dimension reduction parameter dfB ∈ N +; second dimensionality reduction parameter ds＝if(m,c),m<c<(b×n)。
5. The stock recommendation method of claim 1, wherein pre-training each pre-stored deep text matching model to select a target model based on the historical representation of each stock and the stock-selection target feature matrix comprises:
according to the historical portrait of each stock and the stock-selecting target characteristic matrix, pre-training each pre-stored deep text matching model by combining all hyper-parameters of the deep text matching model;
in the pre-training process, evaluating each hyper-parameter combination according to the model performance to obtain a hyper-parameter combination evaluation value set of each model;
calculating a first standard deviation of each model according to the hyper-parameter combination evaluation value set of each model to obtain a first standard deviation set of all models;
and calculating a second standard deviation according to the first standard deviation set, and selecting a target model according to the second standard deviation.
6. The stock recommendation method of claim 5, wherein selecting a target model based on the second standard deviation comprises:
judging the second standard deviation;
if the second standard deviation is smaller than a preset first threshold, taking a model corresponding to the minimum standard deviation in the first standard deviation set as a target model;
if the second standard deviation is greater than or equal to a preset first threshold and less than or equal to a preset second threshold, taking a model corresponding to the maximum standard deviation in the first standard deviation set as a target model;
and if the second standard deviation is larger than a preset second threshold value, taking a model corresponding to the middle standard deviation in the first standard deviation set as a target model.
7. The stock recommendation method of claim 5 or 6, wherein training the target model based on the historical representation of each stock and the stock-selection target feature matrix comprises:
selecting a hyper-parameter combination with the largest evaluation value from the hyper-parameter combination evaluation value set of the target model as the optimal hyper-parameter combination of the target model;
and training the target model according to the optimal hyper-parameter combination.
8. A stock recommendation device, comprising:
the acquisition module is used for acquiring historical stock information of the target stock group;
the stock portrait module is used for extracting the stock portrait characteristics according to the historical stock information so as to obtain the historical portrait of each stock in the target stock group;
the stock selecting characteristic extracting module is used for calculating the corresponding stock selecting target characteristic of each stock according to the historical portrait of each stock and the historical earning rate of each stock so as to obtain a stock selecting target characteristic matrix of the target group of stocks;
the model selection module is used for pre-training each pre-stored depth text matching model according to the historical portrait of each stock and the stock selection target characteristic matrix so as to select a target model;
the model training module is used for training a target model according to the historical picture of each stock and the stock selection target characteristic matrix;
the acquisition module is also used for acquiring the forecast time and the historical stock information of the stock group to be recommended, and inputting the forecast time and the historical stock information of the stock group to be recommended into the target model so as to obtain a sorting result of the stock group to be recommended at the forecast time based on the stock-selecting target characteristic.
9. A computer device comprising a memory, a processor, and a stock recommendation program stored on the memory and executable on the processor, the processor implementing the stock recommendation method of any one of claims 1-7 when executing the stock recommendation program.
10. A computer-readable storage medium, on which a stock recommendation program is stored, the stock recommendation program implementing the stock recommendation method of any one of claims 1 through 7 when executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010357317.6A CN113570455A (en) | 2020-04-29 | 2020-04-29 | Stock recommendation method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010357317.6A CN113570455A (en) | 2020-04-29 | 2020-04-29 | Stock recommendation method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113570455A true CN113570455A (en) | 2021-10-29 |
Family
ID=78157702
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010357317.6A Pending CN113570455A (en) | 2020-04-29 | 2020-04-29 | Stock recommendation method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113570455A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116681522A (en) * | 2023-06-05 | 2023-09-01 | 深圳价值网络科技有限公司 | K line form stock selection method and device, terminal equipment and storage medium |
-
2020
- 2020-04-29 CN CN202010357317.6A patent/CN113570455A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116681522A (en) * | 2023-06-05 | 2023-09-01 | 深圳价值网络科技有限公司 | K line form stock selection method and device, terminal equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Souma et al. | Enhanced news sentiment analysis using deep learning methods | |
| CN117787569B (en) | Intelligent auxiliary bid evaluation method and system | |
| CN111738856A (en) | Stock public opinion investment decision analysis method and device | |
| Bai et al. | Leveraging generative artificial intelligence for financial market trading data management and prediction | |
| Yang et al. | Deep learning model-driven financial risk prediction and analysis | |
| Zhang et al. | Predicting stock price movement using a DBN-RNN | |
| Singhal et al. | Data driven analysis for startup investments for venture capitalists | |
| Ben Hamadou et al. | Forecasting Bitcoin returns using machine learning algorithms: impact of investor sentiment | |
| Jin | GraphCNNpred: A stock market indices prediction using a Graph based deep learning system | |
| CN113570455A (en) | Stock recommendation method and device, computer equipment and storage medium | |
| Ruan et al. | Deep Learning Based on Hierarchical Self‐Attention for Finance Distress Prediction Incorporating Text | |
| CN114529063A (en) | Financial field data prediction method, device and medium based on machine learning | |
| Ma et al. | A Deep Choice Model for Hiring Outcome Prediction in Online Labor Markets | |
| Van Pham et al. | Applications using hybrid intelligent decision support systems for selection of alternatives under uncertainty and risk | |
| Tayal et al. | AI Based Stock Market Prediction-Theoretical Perspective | |
| Ansary | Machine Learning for Predicting the Stock Price Direction with Trading Indicators | |
| Kumari et al. | Brain Tumor Detection Using AIML | |
| Zhang et al. | Graph-Based Stock Prediction with Multisource Information and Relational Data Fusion | |
| TANKSALE | A Data Driven Approach to Option Pricing | |
| Shahani et al. | INVESTOPAL-Smart Financial Investment Advisory System | |
| Sharma et al. | Deep Learning Tools for Predicting Stock Market Movements | |
| Sauchanka | Algorithmic portfolio management: markov chains case | |
| AlHakeem et al. | Iraqi Stock Market Prediction Using Hybrid LSTM and CNN Algorithm | |
| Ferrouhi et al. | A comparative study of ensemble learning algorithms for high-frequency trading | |
| Soukhtanlou | European Union Societal Resilience Prediction |
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: 20211029 |