15 Traffic management and Transportation. This project will aim

15 Jan 2018 Trinity College DublinM.Sc. in COMPUTER SCIENCEResearch ProposalStudent NameAkash LakshmanEmail [email protected] SupervisorMelanie BourocheStudent ID17312220StrandData ScienceDissertation TopicDemand Prediction and Optimal Routing of Dockless BikesRESEARCH AREAThis project predominantly falls into the Smart Cities category. Big cities are now trying to fulfil exceeding demands of the people with respect to Traffic management and Transportation. This project will aim to incorporate the areas of improvement on one such means of transport and that is Dockless Bikes. The ideal aim is to predict the demand for Bikes based on various impacting features. This project will require research in areas such as Predictive Analytics, Deep Learning and elements of machine learning. Probably by finding an optimal approach to execute the task in hand the project will require experiments with different algorithms and ensembles. Furthermore, Weather and Traffic analysis is also another Area that will be concentrated on for Optimal Bike routing purposes.RESEARCH QUESTION1. Considering the advancement of smart city concept Dockless bikes, would it be feasible to predict the demand for bikes depending on all the factors that may influence such a demand via predictive models, regression techniques and ensembles and application of deep learning.2. Further, what could be the factors that influence the generation of optimal routing for a given source and destination. Can machine learning contribute to understanding these factors so as to generate ideal routes based on historical data.RESEARCH SUMMARYFocus on generating bike sharing demand is an extremely innovative idea to increase revenue for a company and service for customers . Dockless bike sharing is a concept that is part of the Smart City innovations that have been trending off late. With the rapid inflation in the traffic due private transport and the pollution level rise leading to a lethal environment many agencies have been developing innovations to solve such issues and reduce the hazards that inevitably occur otherwise.Dockless bikes is a system where bikes are available to share by individuals for temporary basis to get from a source to a destination. The bikes are application based, comprise of a gps locator in the bike and the movement is tracked with help of the application. Now that the concept is implemented next thing to do is how to make its demand optimal as the bike lease demands may vary based on various factors. Further predicting the optimal routes for the user to take shall also be an added improvement.15 Jan 2018 Trinity College DublinThe following table shows roughly the data that will be considered to carry out this task. The data will need to be pre-processed to a usable format of course.Apart from this given data other information such as weather API, traffic information, time of day, day of week, month of the year, etc shall also be considered for the research. The research aims to implements to find the most optimal techniques that can be used to generate the demand. Deep learning Neural Nets, Ensembles can contribute to achieve the goal. The research also hopes to generate correlations and visualize this data so as to understand influencing factors and study patterns and performance of the models. The prediction could be basis of a given amount of time for a given location on basis of historic data.Based on the description the project will aim to provide the customer which route to take and how many bikes are available near his current location that can be used to rent. More importantly the bike providers will be informed the amount of bikes that will be required at a given location.RESEARCH MOTIVATIONThe keen interest on data and the analytics that can be performed was the primary reason I took up a course like Data Science for my masters. This project enables me to pursue exactly that. Furthermore, the smart city concept is trending and there are a lot of developments that are taking place in this field. Therefore, I felt contributing to this domain would be an interesting thing to pursue. After coming to Dublin the Dublin Bikes system got me completely interested, so Urbo Solutions seemed like an extension to what Dublin Bikes is able to achieve. Popularising such an idea could have an impact on many countries as well enforcing reduction in fuel usage. Cleaner and Unpolluted environment is certainly a motivating goal to aim for. With this dissertation I will be able to deep dive into studying data and get a platform for applying a lot of concepts that has been taught in the masters course thereby understand the domain knowledge to a better extent.RESEARCH AIMSThe key ambition of this research is to understand the data of a Dockless Bike agency, which is a bike sharing forum in light of smart city developments. Further, the research aims to deduce the demand prediction with respect to the usage based on multiple contributing factors. The project will try to15 Jan 2018 Trinity College Dublinconsider all the impact that can enhance the bike demand such as the area, day of the week, time of day, time of year, etc.Another objective for this research is to present the user with optimal routing. Traffic is a problem almost every major city in the world is facing. So along with the demand prediction if the user can get the fastest/easiest route to travel from the source to destination that would be very handy, time saving, safe. Here to all the influencing factors such as Weather conditions, traffic report, distance, etc shall be considered.From a technical perspective the research will aim to use the historic data to therefore use various regression techniques find the optimal predictions to generate the demand. Simultaneously machine learning techniques shall be used to understand the routing from a given source to destination so as to get optimized routes going forward. The research will use the data from Urbo Solutions, weather API, Traffic information and load them into the project.The research will involve study of the state of the art from multiple citations and work in this area. It will further highlight what is currently done and what needs to be done in this field ie., the bottlenecks currently present. For this the respective data will be studied thoroughly so as to understand the influence factors in the data.Multiple models shall be made via techniques of regression, ensembles, neural nets and thereby, cross-verified to highlight the most optimal way of achieving the goal and projecting any shortcomings that may be present for future enhancements.POTENTIAL BENEFITS OF STUDY FOR THE FIELDThe benefits would be that there will be satisfied users who could avail the bikes as per their will and locations. Issues such as unavailability of bikes at a particular location is a problem Dublin Bikes currently face. This issue can be solved. Further the users can get from source to destination in the most convenient manner.Conditions such as weather, traffic have not been formerly implemented and considered to generate demand predictions.ETHICAL ISSUESThe company Urbo Solutions will be sharing the data for the dissertation. The company would be sharing data that has already been censored for my perusal.EVALUATION CRITERIONThe research shall be evaluated on the basis of whether an actual demand based on all the factors considered for this dissertation is actually predicted. The accuracy and error will play an important role in determination of the same.SOURCES OF LITERATURE15 Jan 2018 Trinity College DublinGoogle Scholar does have many articles in this domain. Furthermore, there are many Dockless Bike companies based in other countries such as USA, China, etc where information for this research would be useful.Some links are as follows:> http://urbosolutions.com/> https://www.citibikenyc.com/> https://www.bicycling.com/culture/chinese-dockless-bike-share-graveyard> Kaggle, Analytics Vindhya, IEEE> https://rpubs.com/jk100a/MATH664EDPapers & Articles:https://arrow.dit.ie/cgi/viewcontent.cgi?referer=https://www.google.ie/&httpsredir=1&article=1083&context=scschcomdishttps://www.google.ie/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&cad=rja&uact=8&ved=0ahUKEwjOi-bwvM3YAhWnDcAKHb_GDDMQFggvMAE&url=http%3A%2F%2Fwww.mdpi.com%2F2076-3417%2F8%2F1%2F67%2Fpdf&usg=AOvVaw2pg0B6QMwLxklVO0NjzS1fhttps://cseweb.ucsd.edu/classes/wi17/cse258-a/reports/a114.pdfhttp://cs229.stanford.edu/proj2014/Jimmy%20Du,%20Rolland%20He,%20Zhivko%20Zhechev,%20Forecasting%20Bike%20Rental%20Demand.pdfhttps://cseweb.ucsd.edu/classes/wi17/cse258-a/reports/a050.pdfhttp://efavdb.com/bike-share-forecasting/PERSONAL EXPERTISEApart from the knowledge the MSc in Data Science is providing I do have an analytical background which would be influential in this research. I have previously worked to improvement performance better of Data Analytics in Banking and Insurance domain. A decent knowledge in R and python should definitely help in pursuing this project. Prior knowledge of machine learning would also be beneficial for optimal routing generation.COLLABORATIVE PROJECTSMy research is an extension to the developments taking place in the company called Urbo Solutions. This company aims to upgrade their existent system to ultimate generate a better service to the customers and as well as revenue for themselves. Therefore, they are willing to implement any innovation that is interesting from their perspective. They wanted to primarily predict the demand generated, alongside find solutions for bike safety, handling bad parking, revenue improvement. There are other students pursuing the other bottlenecks that can be encountered in this company in this business.15 Jan 2018 Trinity College DublinREFERENCES1. https://arrow.dit.ie/cgi/viewcontent.cgi?referer=https://www.google.ie/&httpsredir=1&article=1083&context=scschcomdis2. https://cseweb.ucsd.edu/classes/wi17/cse258-a/reports/a114.pdf3. http://cs229.stanford.edu/proj2014/Jimmy%20Du,%20Rolland%20He,%20Zhivko%20Zhechev,%20Forecasting%20Bike%20Rental%20Demand.pdf4. https://cseweb.ucsd.edu/classes/wi17/cse258-a/reports/a050.pdf5. http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7959977&isnumber=79599516. https://www.aaai.org/ocs/index.php/WS/AAAIW15/paper/viewFile/10115/10185MILESTONES TABLE