PROFIT PREDICTION USING REGRESSION MODEL FOR TRAVEL AGENTS

• Main Author: PRATAMA SANTI (NIM: 23516035), RAHMATIKA
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/30144
• Institution: Institut Teknologi Bandung

<p align="justify">The Increasing of public interest in the air transport by aircraft occurs year by year, so this opportunity can be exploited by travel agents to improve transactions and corporate profits. The increase is proportional to the number of transactions from the sale of flight tickets to conventionally processed by travel agents and is not used anymore. Aircraft sales history data from various airlines and destinations stored over the years can be described, identified factors that affect profitability, and make predictions. <br /> <br /> The transaction data needs to be preprocessing stages in order to proceed to the next stage. Description of the company's business conditions, can be done by using data mart and visualization. While to identify the factors that affect the company's profit is identified by using correlation. Correlation is not only used to identify the strength of the relationship, but can also be used as a method for the selection of attributes to be used to make predictions. Prediction of profit is done by using data mining that is regression method. Three regression methods used are linear regression, multilayer perceptron, and M5 model trees (M5P). The three techniques used to build the best model for prediction. The resulting model will be evaluated by k-folds cross validation, with k value being 10. From some existing models, will be eliminated based on Root Relative Square Error (RRSE) contained in each model. The model with the smallest RRSE will be the best model for the prediction of air ticket sales data. <br /> <br /> The results of the visualization provide an overview of the variables that can be utilized to increase the profit of the airline, departure time, netCommission, and markup. Based on the correlation result, there are four variables that have strong and positive correlation coefficient, ie markup, netcom, net price, and unit with correlation coefficient between 0.65 to 0.83. The experiments performed show that the profit of the firm is affected by the markup, net commission, and the net price of the airfare. The four variables then form the best prediction model of the ticket sales transaction data, generated by multilayer perceptron technique with the smallest RRSE value is 4.63%.<p align="justify">