Chakrabarty, AmitabhaMahmud, Booshra NazifaFerdoush, ZannatulMim, Lamia Tasnim2019-07-142019-07-1420192019-04ID 15301020ID 15301068ID 15301052http://hdl.handle.net/10361/12353Cataloged from PDF version of thesis.Includes bibliographical references (pages 45-49).This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science and Engineering, 2019.Bangladesh being one of the ve fastest growing economies in the world with its enormous 164.7 million population is facing a huge challenge of adapting to the surging demand of electricity rising throughout the country to support its blooming economy. Load forecasting can play a vital role to overcome this challenge as it serves as an imperative tool behind electric utilities planning and operation management. Forecast loads serve as the basis of many operational decisions such ensuring maximum utilization of power by avoiding under or over generation; understanding future load demand to make economically viable investment decisions; management of resources; infrastructure development along with maintenance schedule planning. This study, rst of all, proposes an automated model that fetches data from daily load generation reports (kept in pdf format) found in the Bangladesh Power Development Boards website [54] to generate a compact dataset of our historical load data with which we have conducted this research. The necessity of this model is no publicly available dataset have been found so far that contains the historical load data along with data of other important features that e ect the forecast load. Secondly, we have approached three major machine learning methods { K Nearest Neighbor, Random Forest, and Long Short Term Memory (LSTM) to observe how these algorithms perform in forecasting load based on the historical electric load data of Bangladesh and what features play important roles to accurately forecast load. We have found that among these three algorithms; LSTM yields the best result having minimal prediction error compared to the other algorithms.49 pagesenBrac University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission.Load forecastPredictionDecision treeRandom forestK NearestLong Short Term MemoryMachine learningDecision treesThesis