Journal of Contemporary Issues in Business and Government Vol. 27, No.5,2021 https://cibg.org.au/ P-ISSN: 2204-1990; E-ISSN: 1323-6903 2385 FORECASTING USING NEURAL NETWORKS AND STOCHASTIC MODELS ON DAY OF THE WEEK EFFECT: A CASE STUDY OF KSE 100 INDEX Azhar Ali Marri 1,2 and Mir G.H.Talpur 1 1 Department of Statistics, University of Sindh Jamshoro Pakistan 2 Department of Statistics, University of Balochistan Quetta Pakistan Abstract The main objective of this research was to determine the day of the week effect in Karachi stock exchange (KSE) 100 Index. The problems of financial market structure is analyzed and forecasted by many statistical models. For example Auto regressive integrated moving average (ARIMA) model and artificial neural network (ANN) models. In this research day of the week effect was investigated Wednesday found significant and Monday was noted not significant. 15 step ahead prediction of Wednesday was observed through ARIMA model and ANN models. The coefficient of correlation for actual and forecasted values was perceived 0.8871 by ARIMA model and 0.924 by ANN model. Power of accuracy was displayed 88.765% by ARIMA model and 97.18% by ANN model. Keywords; Anomaly, Wednesday, KSE 100 Index, ARIMA, ANN 1. Introduction The time series is chronologically observed historical data including large in size high dimensionality and essential to update continuously. The cumulative usage of time series observations are initiated with a great deal of investigation and improvement to endeavor the time series study (Kumar and Murugan 2013). Time series modelling and prediction is a significant area of research where past recorded values of the same variable are collected and investigated (Zhang 2003). The prediction of the future event based on present and past observable events (Yao and Tan 2001). Numerious studies are reported in the literature for stock exchange prediction and it is still an active part of the study (Adebiyi, et al. 2012). Day of the week effect has attracted considerable attention since its discovery back in 1930 (Gharaibeh and Hammadi 2013). An adequate number of studies are conducted on day of the week effect. Ko, Li and Erickson (1997) reported that stock prices fluctuates with day of the week. Aly, Mehdian and Perry (2004) evaluated variation in daily stock exchange prices. Rossi and Gunardi (2018) examined calander anomaly (CA) and found recurring anomalies in stock market. The stock exchange prices prediction acquired attention of many researchers for private and institutional sectors. Selvan and Arun (2012) reported stock exchange prices are highly irregular with time and generally follows nonlinear pattern. Adebiyi, Adewumi and Ayo (2014) forecasted ANN and ARIMA model using New York stock exchange (NYSE) daily closing price index. Emin (2007) forecasted daily and seasonal data of IMKB 100 Index with neural network models. Saiful and Yoshiki (2011) examined the applications of ANN model for forecasting of mudharabah time credit return. Olatunji, et al. (2011) predicted Saudi stock market with ANN model and observed lowest forecasting error with coefficient of correlation up to 99.9%.