Please use this identifier to cite or link to this item:
|Title:||Using graph trends to predict short term changes in the stock market|
|Abstract:||The stock market is prone to frequent small fluctuations which could be used to the advantage of stock market traders where both the profit and loss of purchased and sold stocks is multiplied by the leverage. This project aims to build a database of past market data for ten stocks, five from Internet companies such as Facebook and Twitter, and the other five from sales-based companies like Apple and Microsoft. The effect each particular grouping would have on the classification of a stock was also taken into consideration, considering stocks individually, then considering all internet companies as one group and production companies as another, and finally all companies will be compared together. A software solution using neural networks and mathematical calculations based on technical data, or data generated from the stock’s graph, was built in order to predict whether a stock will rise, fall or stay stable after a five day period. The data was also stored and analysed using a data-mining software in order to better understand the workings of the market and be able to represent this in a graphical format. As expected data from RapidMiner was more accurate and did show a correlation between a number of factors which could lead to a rise or fall in the market. In this project, past data was gathered for 10 individual stocks, this data was then analysed to look for patterns. An Artificial Neural Network was created which took the nine most important stock factors as inputs and provided an output between 1 (Buy) and 0 (Sell). The results were positive yielding an overall average accuracy of 61%, with some stocks being predicted to 78% accuracy. This shows clearly that there is a relation between technical data and short term price fluctuations which can be used for predictive purposes.|
|Appears in Collections:||Dissertations - FacICT - 2016|
Dissertations - FacICTCIS - 2016
Files in This Item:
|4.53 MB||Adobe PDF||View/Open Request a copy|
Items in OAR@UM are protected by copyright, with all rights reserved, unless otherwise indicated.