Development of passenger train service quality model for special occasion through neural networks and fuzzy inference system

Abstract

The Adaptive Neuro-Fuzzy Inference System (ANFIS) and an Artificial Neural Network (ANN) namely Probabilistic Neural Network (PNN) techniques were used in this thesis to model intercity train passengers’ perception on its service quality (SQ). A stated preference survey was carried out with 6 skilled enumerators of intercity train users at Kamlapur Railway Station, Dhaka on the month of July, 2016. There are three sections in the survey questionnaire. The first section aims to get demographic and socio-economic information (age, gender, occupation etc.) of commuters and the reason for using intercity trains. The second section focuses on 18 attributes that are accountable for the evaluation of intercity train SQ. The third section organized to get priority ranking of the attributes from the respondents. These attributes were in a close ended arrangement with relevant multiple choices. The respondents were asked to assess the present situation of the service by marking the checkboxes from their point of view against each attribute. The multiple-choice check boxes are numbered by 1 to 5 where “5” corresponds to excellent quality and “1” corresponds to very poor quality. After survey, incomplete data sets were screened out from collected data. Finally, 1037 and 553 user’s data were used to calibrate the ANFIS and PNN structures for intercity train SQ estimation during regular days and special days, separately. The training and forecasting sets contained 80% of whole sample (830 samples for regular days, and 443 samples for special days) and 20% of whole sample (207 samples for regular days, and 110 samples for special days) observations, respectively. MATLAB 2014b is used for the development of these models. The proposed ANFIS structures with eighteen attributes showed 54.1% and 60.2% accuracy and PNN structure showed 50.7% and 57.3% accuracy in predicting train SQ for regular days and special days, respectively. Finally, a stepwise approach was followed for ranking the intercity train SQ attributes influencing its overall SQ and the results were compared with that of the empirical observations (public opinions). Study found that besides waiting place condition, attributes related to physical conditions and service features of intercity train are important determinants of its perceived SQ for regular days and special days, respectively. Beside waiting place condition, ‘Toilet cleanliness’, ‘Fitness of car’, ‘Air ventilation system’, ‘Convenience of online ticketing system’, ‘Seat comfort’, ‘Ease at entry and exit’, were the most significant physical attribute those influence the users’ decision-making process on regular days. In contrast, on special days, ‘Travel cost’, ‘Air ventilation system’, ‘Convenience of online ticketing system’, ‘Car arrangement’, and ‘Travel delay’ were the most significant service attribute which influence the users’ decision making process.