STOCK MARKET TREND PREDICTION USING MACHINE LEARNING ALGORITHMS
Loading...
Date
2018-11
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UMT, Lahore
Abstract
In financial world stock price prediction is an important issue. Even minute predictive
performance progress can be extremely profitable. This thesis deals with difficulty of predicting
the trend of movement of stock index for Pakistan stock exchange. The rationale of present
study is to set standard for ensemble machine learning method (random forest algorithm)
against single classifier model (decision tree algorithm) to find out best accuracy ratio for
prediction. These machine learning models are used with three approaches for input; the first
approach includes eight technical indicators calculated from high, low, open, close values of
Karachi stock exchange 100 index as input variables, second approach includes three
fundamental indicators and lastly third approach encounter both technical and fundamental
indicators as input variables. Models evaluation is carried out from 2010 to 2017 time period
with 80% training data and 20% test data. Experimental results revealed that in all three input
approaches random forest algorithm outperform decision tree algorithm with 75.45%, 57.31%
and 76.55% accuracy ratio respectively. Results also show that prediction performance of both
algorithms increases when technical-fundamental input approach apply on them. This study
plays a part to literature as to the best of our information and facts, the initial to make such a
huge benchmark for KSE 100 index data.