Browsing by Author "Islam, Md.Touhidul"

Now showing 1 - 2 of 2

A Study on Public Bus Fare Elasticity for Different Routes in Urban Area
(Daffodil International University, 22-04-24) Islam, Md.Touhidul; Rahman, Mohammad Sarwar; Shimul, Nousad Islam
There is a long debate regarding public bus fare in Bangladesh, that, the bus fares are not allocated according to passengers’ opinion, rather than, bus operators decision dictates. Though Government have fix designated bus fare inventory, those are little practiced in the field. Therefore, issues of bus fare remain questionable and seem to be unjustified. A shed in public opinion regarding bus fare is come to a focus in this regard. In this study, we design a question survey to investigate bus fare sensitive with respect to bus trip among passengers. We perform survey in two routes of Dhaka city. The route 1 is Mirpur to Dhanmondi and Route 2 is Dhanmondi to New Market. In total 101 respondents take part in the survey. Among them 57.43% were male within the responses. 42.57% respondents had monthly household income within the range of 30-50k BDT. Large number of respondents do not have personal vehicle. Among the collected data, 73.27% respondents were travelling in peak hour. Almost 82% respondents considered that existing bus fare is not expensive. 5.94% respondents were captive rider, i.e., not willing to change modes at all. The demand curve, in this case, %bus fare increase vs. %ridership curve, show that doubling bus fare leads the curve to the infeasible region, where ridership reduces drastically. The Model 1- Linear model and Model 2- Polynomial model have degree of determinacy R2 value 0.89 and 0.97 respectively. Route 1, i.e., Mirpur to Dhanmondi is more sensitive to fare increasing compare to the other counterpart. Male are more sensitive to fare increment. At low or high fare increment, both male and female react similarly. 31-50 years age passenger groups are most sensitivity. Elderly and very young passengers are less sensitive to higher and lower bus fare increment respectively. Low income and high-income respondents are very sensitive and less sensitive to bus fare increment respectively. Off peak hour passengers are more sensitive to bus fare change. Slight increase in bus fare affects ridership more for non-AC local bus passengers. At lower fare increment, seating bus passengers’ ridership remain similar regardless of AC or non-AC buses. Fare elasticity at 30% increment level is -0.11, that is, increase in 1% additional fare will decrease ridership by 0.11%, which is inelastic. At 70% fare increment level, fare elasticity for male is -1.09, that is elastic. The increase in 1% additional fare decrease ridership by 1.09%. Fare elasticity for 18-30 years age group at 30% fare increment level is 0.0, that is perfectly inelastic. Increase or decrease of additional fare will not change ridership at this level. Fare elasticity for non-AC local bus passenger group at 70% fare increment level is -1.8, that is inelastic, increase of 1% additional fare will decrease ridership by 1.8%. The fare elasticity investigation can be extended for large scale survey by concerned authority. Future bus fare adjustment should be based on public survey and bus fare sensitivity among different groups of passengers.
Authentication of bank notes and image processing through Artificial and Convolutional Neural Network
(Daffodil International University, 2018-05-07) Islam, Md.Touhidul
This paper focuses on two common applications of image processing done with the help of neural network. At first, it gives a solution for detection of counterfeit bank notes using Artificial Neural Network with the structured image dataset of the corresponding bank notes provided by the UCI Machine Learning Repository. In the second part, the emphasis is on solving the unstructured image classification problem through Convolutional Neural Network and using the famous Kaggle dataset. Both problem include a common Artificial Neural Network architecture. However, in the second problem, on top of Artificial Neural Network, two convolutional layers are added. Two classifications discussed here are non-linear in nature and they have some inherent differences. This paper also discuss the common pitfalls and possible improvement techniques in the future implementation of this strategy along with a detailed discussion.