TW202213239A - A predicting the trend of stock prices and trade advising system based on neural networks of analyzing multiple technical analysis indicators - Google Patents

Abstract

A predicting the trend of stock prices and trade advising system based on neural networks of analyzing multiple technical analysis indicators is disclosed. The system at least comprises: user interface to input targeted stock and receive the output; data gathering system to gather the historical data on stock; one data extension system to calculate multiple technical analysis indicators which as one part of the train data for neural networks; a neural network system analyzes the multiple technical analysis indicators and historical data to predict stock price before trading; and a trade advising system to advise users to trade by the stock prices trend and the specified technical analysis indicator.

Description

A neural network-like analysis of multi-technical indicators for stock price trend prediction and transaction decision-making assistance system

The present invention relates to a stock price prediction, trend prediction and transaction decision assistance system in the securities trading market. After collecting transaction-related information, after generating a plurality of relevant technical indicators, based on a neural network, according to market transaction information and a plurality of After analyzing the technical indicators, it is a system that predicts the stock price, analyzes the stock price trend, and provides trading decision suggestions before the market opens.

At present, there are many technical indicators in stock market analysis, such as Stochastic Oscillator, Relative Strength Index (RSI), Williams %R, Directional Movement Index (DMI), smoothing Similarities and Differences Moving Average (Moving Average Convergence Divergence; MACD) Bollinger Bands, etc., but in auxiliary trading or forecasting stock prices, there are situations that cannot be accurately analyzed. For example, the stochastic indicator will passivate, that is, the indicator has shown the current High point or low point, but it is impossible to know whether it will continue to rise or fall in the future, and it needs to be judged by other indicators, and there is a conflict between indicators, that is, two or more indicators show different forecast trends, and it is difficult to judge the rise or fall ; There are also different trading characteristics of each stock, which makes the indicators suitable for analysis also different, and because of the need to be familiar with a variety of technical indicators, it is easy to make mistakes in manual analysis.

Trading decisions based on technical indicators of stock market analysis, there are also indicators Passivation, conflict between indicators, and the use of different indicators for different trading characteristics of each stock make trading decisions and judgments inaccurate and lose profit opportunities.

In view of the above problems, the present invention constructs a stock price prediction, trend analysis and transaction decision assistance system with a neural network, alleviates the problems of passivation of technical indicators and conflicts between technical indicators, and establishes and analyzes historical transaction related information (opening price, highest price, lowest price) Price and closing price, etc.) and the values of multiple technical indicators, predict the stock price and stock price trend before the market opens to provide judgment, and provide trading (buy, sell) decision-making suggestions with the help of common technical indicators of users.

The present invention uses the historical transaction information of individual stocks to automatically extend and calculate various technical indicators, so that the neural network system can learn from multiple technical indicators and historical transaction information as training data, and the present invention automatically assists in judging other technical indicators, reducing the use of It takes time for users to learn indicators, and users only need to be familiar with 1~2 technical indicators.

The analysis system of the present invention uses a neural network to connect historical transaction-related information, multiple technical indicators, etc. as learning materials, and can automatically adjust and apply each technical indicator to the analysis of individual stock prices. Predict the stock price before the market opens based on recent data as the basis for analysis.

The present invention includes a data collection system for collecting historical transaction-related information of individual stocks to be analyzed. The historical transaction information of individual stocks at least includes historical opening price, highest price, lowest price, and closing price, etc., and various historical transaction values in the stock market can be Historical data objects are stored in the historical data database.

The present invention includes a data extension system, which automatically calculates and generates associated data with the opening price, the highest price and the lowest price after retrieving data from the historical data database, The associated data includes at least the currently used technical indicators, Stochastic Oscillator, relative strength indicator, Williams indicator, trend indicator, smoothed moving average of similarities and differences, Bollinger Bands, etc., and stored in the basic data database. The data extension system has Future expansion can be expanded after the development of new technical indicators.

The invention includes a neural network stock price analysis system, which uses data database data as the learning data of the neural network, and automatically establishes the correlation weight between each value of the basic data database and the historical stock price, and based on the basic data database before the opening of the market Predict the highest price, lowest price, closing price and ups and downs of trading.

The invention includes a stock price trend prediction system, which predicts the stock price up, down or flat according to the pre-market trading highest price, lowest price, closing price and ups and downs predicted by the neural network stock price analysis system.

The present invention includes a decision-making assistance system, based on commonly used stochastic indicators, Bollinger channels, etc., for example, when KD crosses or the stock price hits the upper or lower trajectory of the channel, it is supplemented by a stock price trend prediction system to analyze the trend and slow down the Solve the situation of inaccuracy due to passivation of technical indicators alone, and output a decision to buy or sell.

10: User Interface

20: Neural network-like analysis of multi-technical indicators for stock price trend prediction and transaction decision-making assistance system

21: Data collection system

22: Historical Data Repository

23: Data extension system

24: Basic Data Repository

25: Neural network-like stock price analysis system

26: Stock price trend prediction system

27: Trading Decision Aid System

FIG. 1 is a system architecture diagram of the present invention.

In order to facilitate the understanding of the technical features, content and advantages of the present invention and the effects that can be achieved, the present invention is described in detail with diagrams and in the form of embodiments, and the diagrams used are only for illustration and auxiliary description. , which may not necessarily correspond to the real proportions and precise configurations of the present invention after the implementation of the present invention. Therefore, the relationship between the proportions and configurations shown in the accompanying drawings should not be interpreted, but to limit the present invention to the actual implementation. right.

Please refer to FIG. 1. FIG. 1 is a diagram showing the structure of each module of the stock price trend prediction and transaction decision assistance system of a kind of neural network analysis of multi-technical indicators according to an embodiment of the present invention. The overall system includes a user interface 10 and a Neural network analysis multi-technical indicators stock price trend prediction and transaction decision assistance system 20 . The user interface 10 is used for inputting individual stocks to be analyzed and receiving system output results. The transaction decision assistance system 20 includes a data collection system 21 , a historical data database 22 , a data extension system 23 , a basic data database 24 , a neural network-like stock price analysis system 25 , a stock price trend prediction system 26 and a transaction decision aid system 27 .

According to an embodiment of the present invention, the data collection system 21 collects various historical transaction related information of the individual stocks to be analyzed, including at least historical high price, low price, opening price, transaction price, etc. The input method can be input by the user or automatically. Collect and store the data in the historical database 22 as basic data for future calculations.

According to the embodiment of the present invention, the data extension system 23 will automatically calculate various technical indicators after retrieving data from the historical database, at least including but not limited to stochastic indicators, relative strength indicators, Williams indicators, smooth moving average of similarities and differences, trends Indicators and Bollinger Bands, etc., and retain the design of new indicators in the future, extend the information related to stock prices, strengthen the characteristics and correlation of historical stock price information and various historical related information, and store historical transaction related information and extended data into basic data Database 24.

According to the embodiment of the present invention, the neural network-like stock price analysis system 25 uses the basic data database 24 as training data, and the basic data database at least includes but is not limited to historical transaction related information, stochastic indicators, Bollinger channels, Williams index, relative Strength indicators, trend indicators, etc., can be added and expanded, and used as training data, and the historical highest price, lowest price, The closing price and the ups and downs are used as verification data, and the neural network is used for correlation and training to establish the correlation between each indicator value and historical transaction information, and to analyze and generate predicted individual stock transaction information, users only need to be familiar with 1 to 2 technical indicators As a basic judgment, the system automatically assists in analyzing other input indicators, simplifying the judgment of traders and reducing manual judgment errors.

According to an embodiment of the present invention, the stock price trend prediction system 26 uses the highest price, lowest price, closing price, and ups and downs of individual stocks predicted by the neural network stock price analysis system 25 before the opening of the market as the stock price trend prediction, and the prediction system uses two modes. Stock price trend forecast: general confidence mode and enhanced confidence mode.

General confidence model: the highest and lowest prices of individual stocks predicted by the neural network-like stock price analysis system 25 are both higher than the highest and lowest prices of the previous day’s closing, then the stock price trend forecasting system 26 predicts that the stock price will rise; The highest and lowest prices of a stock predicted by the stock price analysis system 25 are lower than the highest and lowest prices of the previous day’s closing, and the stock price trend prediction system 26 predicts that the stock price will fall; If the trend of the price and the lowest price is inconsistent, the judgment is unknown.

Confidence enhancement mode: The highest, lowest and closing prices of individual stocks predicted by the neural network-like stock price analysis system 25 are higher than the highest, lowest and closing prices of the previous day's closing, and the stock price trend prediction system 26 predicts that the stock price will rise. ; The highest price, lowest price and closing price of a stock predicted by the neural network stock price analysis system 25 are all lower than the highest price, lowest price and closing price of the previous day's closing, and the stock price trend prediction system 26 predicts that the stock price will fall; In the aforementioned case, the judgment is unknown.

According to the embodiment of the present invention, in the transaction decision assistance system 27, the user can select commonly used technical indicators, supplemented by the stock price trend judgment output by the stock price trend prediction system 26, such as The machine indicator value KD crosses. When the K value breaks through the D value from a low level, it is generally considered to be a buy trading signal, but in fact, there is still KD passivation, and the stock price will continue to fall risk. The system 26 judges the stock price trend of the next day, so as to alleviate the bottleneck and risk of passivation of the original technical indicators, and improve the success rate of profit.

10: User Interface

20: Neural network-like analysis of multi-technical indicators for stock price trend prediction and transaction decision-making assistance system

21: Data collection system

22: Historical Data Repository

23: Data extension system

24: Basic Data Repository

25: Neural network-like stock price analysis system

26: Stock price trend prediction system

27: Trading Decision Aid System

Claims (6)

A kind of neural network-like analysis of multi-technical indicators stock price trend prediction and transaction decision assistance system, suitable for an electronic device, at least comprising: 1. A data collection system, which is input by the user or automatically collects information on historical transactions of individual stocks through the Internet. The information collected shall at least include the historical high price, low price, opening price, closing price, etc. of the individual stock, and store the historical data. database. A historical data database, which stores the information obtained from the data collection system and serves as the data transaction interface between the data collection system and the data extension system. A data extension system that automatically calculates various technical indicators after retrieving from the historical data database. The calculated technical indicators include but are not limited to stochastic indicators, Bollinger bands, Williams index, relative strength indicators, trend indicators, etc., and can be expanded at any time New technical indicators, or historical transaction information and stock price related information, and the data after extended calculation is stored in the basic data database. A basic data database is used to store the data generated by the data extension system, and serve as the data transaction interface of the data extension system and the neural network-like stock price analysis system. A type of neural network stock price analysis system, based on the neural network-like learning of various technical indicators in the basic data database, the highest price, lowest price, closing price, and up-down related information of individual stock transactions, and to predict the future transaction of individual stocks. The highest price , the lowest price, the closing price and the change. A stock price trend forecasting system, which judges the future stock price trend of individual stocks according to the future trading highest price, lowest price, closing price and ups and downs of individual stocks predicted by the neural network stock price analysis system. 1. A trading decision-making assistance system, based on the technical indicators familiar to the user, supplemented by a stock price trend forecasting system to make judgments and generate trading suggestions. According to the stock price trend prediction system described in claim 1, the trend analysis is generated by comparing the predicted highest price, lowest price and closing price of a stock with the highest price, lowest price and closing price of the previous trading day. The prediction mode is divided into general: Confidence Mode and Enhanced Confidence Mode. For example, the general confidence model described in Item 2 of the scope of the patent application is that the highest and lowest prices of individual stocks predicted by the neural network-like stock price analysis system are higher than the highest and lowest prices of the previous day's closing, and the system predicts that the stock price of individual stocks will be higher than the previous day's closing price. Rising; the highest and lowest prices of individual stocks predicted by the neural network stock price analysis system are both lower than the previous day's closing highest and lowest prices, and the system predicts that the stock price will fall; such as the individual stocks predicted by the neural network stock price analysis system If the trend of the highest price and the lowest price is inconsistent, the system judges unknown. For example, the confidence-enhancing model described in Item 2 of the scope of the patent application is that the highest, lowest and closing prices of individual stocks predicted by the neural network-like stock price analysis system are higher than the highest, lowest and closing prices of the previous day's closing price. , the system predicts that the stock price will rise; the highest price, lowest price and closing price of the individual stock predicted by the neural network stock price analysis system are lower than the previous day's closing highest price, lowest price and closing price, and the system predicts that the individual stock price will fall ; If it is not the above-mentioned situation, the system judges unknown. According to the transaction decision assistance system described in claim 1, the technical indicators selected by the user are supplemented by the stock price trend predicted by the neural network stock price analysis system to generate trading signals, such as the KD cross or Bollinger channel of the stochastic indicator. When the stock price hits the upper or lower line of the channel, it is supplemented by the stock price trend prediction system to analyze the trend, so as to alleviate the inaccuracy caused by the passivation of technical indicators alone, and output the decision to buy or sell. According to the neural network stock price analysis system described in claim 1, the historical data training and analysis are performed based on the neural network, not limited to deep convolutional network (Deep Convolutional Network; DCN), recurrent neural network (Recurrent Neural Network) Network; RNN), Long/Short Term Memory Network (LSTM) or other neural network framework (Framework) implementation.

Images