Abusing Pattern Detection System for Stock Market

Abstract

Investors participated in wholesale company’s trade among market contributors at particular granted prices. The stock market is the place that fulfills those requirements. However, with the rapid dissemination of new information, maintaining efficient markets are hard to achieve and maintain. Anomaly is taken more important place which can be repeatedly happening or persists once and disappears. Some of them are led to taking profit using this strange behavior. Investors/ need to be well aware of abnormalities in market parameters and they are likely to get into difficulty due to a lack of perceptions of market fluctuations. Detecting manipulations methods need to exist in the stock market and widely used Rule-based patterns as practiced. However, Operators are changing their trading patterns and they are mostly looking for the newest methods to operate stock market behavior. Rule-based or static recognition methods fail to recognize these new maneuver attempts. The main objective of the project is to overcome these challenges by implementing a research methodology to identify these evolving stock fluctuations patterns. Artificial Immune system (AIS) theories are a class of computationally intelligent, rule-based machine learning systems inspired by the principles and processors of the vertebrate immune system. The algorithms are typically modeled after the immune system characteristics of learning and memory for using problem-solving (Coello, 2005). These applications which are based on AIS theories do not include separate training phases. The novelty of this research is having a training phase using domain expertise knowledge. The system was tested based on transaction data collected from Saudi Stock Exchange that is described in the next chapter. The system is designed to detect price or volume abnormality detection using 10 years back archive data by various machine learning methods to get maximum accuracy. Price change, Volume, Turnover, and Trades feature are extracted through the more than 20 labels in the dataset that gives the best feature combination as the conclusion of domain expertise and using wrapper method which are described in the next chapters. Statistical calculations are used to mark abnormality of selected data set with help of domain expertise. The literature review was done to find similar projects which still have limitations and identify a research gap. Some major design issues were identified when the designing phase and getting rid of them by designing solutions for each of them. The final implementation was done by selecting the most suitable techniques to design web applications. The evaluation phase was done based on the conclusion of the domain expertise knowledge.