School of Business and Economics (SBE)

Permanent URI for this community

https://escholar.umt.edu.pk/handle/123456789/182

Browse

Browsing School of Business and Economics (SBE) by Subject "Artificial neural network"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    A Comparison of Different Weather Forcasting Models
    (University of Management and Technology, 2017) Zaheer Abbas
    In this thesis, we studied the performance of different statistical models and compare their forecast accuracy. In particular, we used multiple linear regression (MLR), seasonal autoregressive fractional integrated moving average (SARFIMA), and artificial neural network (ANN). A dynamic non-linear autoregressive (NAR) back-propagation ANN algorithm has been applied to estimate the forecast accuracy. For ANN model, we used moving average (MA) and Holt-Winter exponential smoothing (HW-ES) transformations for pre-processing the data. The monthly data of different weather parameters have been obtained from the Lahore Metrological department, Pakistanto apply the aforementioned models. The results showed that the ANN model with MA transformation of the data has the smallest root mean square error and the highest correlation coefficient for different weather parameters. Thus, ANN outperforms than the rest models in this study and it can be used to efficiently forecast the weather parameters.
  • Thumbnail Image
    Item
    Predicting stock price movement in Pakistan: Comparison between artificial neural network models and traditional linear models: A case of KSE
    (UNIVERSITY OF MANAGEMENT AND TECHNOLOGY, 2015) Muhammad Abubakr Naeem
    Since the establishment of the Karachi Stock Exchange (KSE) in 1993, Pakistan's stock markets have expanded rapidly. Although this rapid growth has attracted considerable academic interest, few studies have examined the ability of conventional financial models to predict the share price movements of stocks. This gap in the literature is significant, given the volatility of the Pakistani stock market and the additional risk arising from the political and regulatory environs of Pakistan, we examine the relative ability of Hybrid Artificial Neural Networks models to forecast stock returns. In this research, the gap is attended by comparing the predictive ability of three linear models i.e. Univariate ARIMA model, Multivariate CAPM model and Multivariate three factor model. Furthermore, we compared these three linear models with the equally specified artificial neural network models containing the same predictor variables. One of the advantages of artificial neural networks is that they relax the model linearity assumption. The analysis conducted is based on the data from Karachi Stock Exchange, containing data for companies from the period of 2010-2014. The results of the research designate that the multivariate models i.e. the dynamic CAPM and the dynamic three factor model surpass the univariate model, that only incorporate lagged returns of stock prices, in forecasting future returns. Additionally, using the artificial neural networks notably enhance the predictive ability using the same predictor variables. Also, irrespective of the linear or non-linear model used, there is significant difference in the forecasting accuracy of univariate ARIMA, multivariate CAPM and multivariate FF3 model.