Artificial Netural Network
Artificial Neural Network (ANN for short)

It is an abstraction and simulation of several basic characteristics such as the human brain or the Natural Neural Network. Since the first practical form appeared in the early 1990s, it has rapidly gained popularity, and has created software systems adapted to artificial intelligence by simulating biological neural networks.

The artificial neural network is a dynamic system with a directed graph as the topology structure established manually, and it processes information by responding to the continuous or discontinuous input.

Features and Advantages of Artificial Neural Networks
Artificial neural networks are an excellent tool to help businesses meet growing consumer demands.
Advant- age
Can fully approximate arbitrarily complex nonlinear relationships
The parallel distributed processing method makes it possible to quickly perform a large number of operations
All quantitative or qualitative information is stored in equal potential with each neuron in the network, which is inherently robust and fault-tolerant.
Can learn and adapt to unknown or uncertain systems
Ability to handle both quantitative and qualitative knowledge
Self- Organization
Self study
Self- Correction
Don't repeat mistakes
The function of artificial neural network
Dedicate data
With its powerful data settlement ability, unforgettable "memory" and extremely fast computing speed, artificial intelligence AI can analyze and discover market laws through deep learning and other methods, and automatically generate trading strategies accordingly.
The AI system will jointly look for trading opportunities from the technical and fundamental aspects.
The AI system adopts a set of Internet keyword capture technology, which can instantly extract tens of thousands of core keywords from a large number of websites in different languages at home and abroad.
The artificial intelligence algorithm is used to distinguish the positive and negative attributes of words, judge the public psychology, and divide the market sentiment into 5 categories. After synthesizing all kinds of information, the AI system will deduce the future trading model.
Artificial intelligence

AI (Artificial Intelligence) is currently known to be applied to recognition systems, expert systems, automatic planning, intelligent search, positive theorems, games,automatic programming, intelligent control, financial transactions, genetic programming, etc.

Artificial intelligence (abbreviation: AI), artificial intelligence is widely used in financial transactions, and artificial intelligence trading systems are increasingly sought after by investors because of their stable profitability and excellent annualized returns. The artificial intelligence trading system makes the traditional trading concepts regularized, variable, serialized, and modeled, and uses computer technology to screen out a variety of "high-probability" nodes from big data, so as to formulate new investment strategies and form a complete set of operating systems. Practical computer programs are automatically executed in real trading.
Self conscious
Data source
Market data Historical transactions News and sentiment Company data Other data
Formula algorithm Neural network diagram Network decision tree Fuzzy logic
Speed Frequency Holding period Size/Lot
Objective Function Performance and Benchmarks
The predecessor of META FX's X-TECH team came from Silicon Valley in North America. They have been engaged in artificial intelligence research and development for 12 years. The 2009 US subprime mortgage crisis has just ended. Mike, who has worked on Wall Street for many years, suffered heavy losses in this round of stock market crash. Laifa Oh founded X-TECH Technology in Silicon Valley, formed an AI team, and began to develop AI algorithmic trading. After more than 3 years of research and development, AI algorithmic trading has been used in stocks, foreign exchange, futures, digital currencies, etc. More than 8 years of real-time operation inspections have realized the application of artificial intelligence technology in the trading field. Compared with simple and practical EA, the profitability of META FX-AI system after using artificial intelligence technology has been significantly improved.
META FX Artificial Intelligence Model
Trend Game Mathematical Quantitative Model
The practical mathematical model to describe market behavior is called quantitative finance. Learning from the chess form method of the ancients game of Go, mastering the chess form is the key to victory. From price judgment to trend judgment, the quantitative change from point to surface can be obtained. Thereby establishing a quantitative trading model for the trend.
Symbolic Logic Model & Fuzzy Logic Model
Symbolic logic is a form of reasoning that primarily involves the evaluation of predicates. Fuzzy logic lends out binary judgment-true-false constraints and allows any given predicate to belong to a set of true-false judgments of varying degrees, which are defined in terms of set membership functions, for example: the input variables of the model may include Any relevant indicators, quantitative indicators, technical indicators, fundamental indicators or indicators of other nature.
Decision tree model
There are two main types of decision trees: classification decision trees and regression decision trees. Classification decision trees are characterized by the fact that the output variable is categorical (eg buy, hold or sell), whereas regression decision trees are characterized by the fact that the output of the characteristic variable is a numerical value (eg -2.5%, 0%, +2.5%, etc.). The nature of the data in the decision tree training set determines the type of decision tree generation. Algorithms for generating decision trees include C4.5 and Genetic Algorithms.
Neural network model
Neural networks are almost the most favored machine learning model for algorithms. Neural networks consist of layers of nodes interconnected between input variables and output variables.
The model is the core of the algorithmic trading system. In order to make the algorithmic trading system more intelligent, the system should save all historical data of errors and select the appropriate internal model according to this change.
Execution model
Responsible for completing transactions determined by the model kernel
Supervised Algorithmic Trading Model
Must use the system's endogenous data as a training set, allowing the model to make self-aware adjustments to the environment
Market analysis model
Perform the collection, sorting and obtaining of reasonable data to analyze the volume and price information of the daily trading profile of a particular security
Neural network model
Manage market quotes based on a single model, balance unspent expected liquidity balances and then compound management models for multiple markets
Operating model
Adaptively increase or decrease the transaction price to deal with the passive attack system of the multi-market