- The Capstone Project will enable students to apply and integrate what they have learnt and give them an opportunity to delve in greater depth, into one or more of the topics covered in the courses. Faculty supervisors will be assigned to guide the students. The capstone project should be completed within 6 months and 10 months for full time and part time student respectively. The total number of hours committed to the project must be at least 182 hours for the entire duration of the project.
- Capstone Projects are available on a competitive basis. To successfully clinch a Capstone Project, students are required to undergo the Capstone Project sponsor’s selection process which may include interviews and assessments. In some cases, Capstone Projects may include an allowance to the student either in the form of stipend or a scholarship for their work on the project, but this will be at the discretion of the company.
Mode of Work
- Students are to work on their Capstone Projects individually, in collaboration with a sponsoring company, under the supervision of an SMU appointed supervisor. In cases where the project scope is large enough to allow for the involvement of more than one student, a maximum of two may work on the project provided each student makes a distinct contribution towards the project.
- Students may be expected to work on-site at the sponsoring company's premises if necessary. This may help them in understanding the business domain, problem definition and even in gaining access to information systems, documents and resources available within the company. The work arrangement may assume the form of an internship with the sponsoring company.
- Strictest confidentiality is maintained between the sponsoring company, the student and SMU supervisor. Prior permission will be sought from the company before the use of any information, in any way, such as for presentation and report purposes.
MITB Students' Capstone Project Posters
Click on the thumbnails below to view some of the incisive Capstone Project posters undertaken by our MITB students, sorted by project completion year.
WHEN 1 + 1 > 2 – HOW MODERN DATA SCIENCE COULD COMPLEMENT ACTUARIAL SCIENCE IN CLAIM COST ESTIMATION
LOK Jun Haur
NG Jian Ming
Dr. Daniel LIN
TOPIC MODELLING AND PREDICTIVE ANALYSIS ON AIRBNB REVIEWS TO DERIVE RELATIONSHIP TO RATINGS
IMPACT OF DIGITAL PAYMENT ON COMMERCIAL BANKS IN CHINA
Dr. Aldy GUNAWAN
NETWORK CONNECTIVITY ANALYSIS OF BUS PUBLIC TRANSPORTATION SYSTEM IN SINGAPORE
Denise Adele CHUA Hui Shan
EXTRACTING COUNTRY SENTIMENT FROM EARNING TRANSCRIPTS
NEO Yi Peng
Dr. WANG Zhaoxia
OUTSOURCING LIFE CYCLE MODEL FOR FINANCIAL SERVICES IN THE FINTECH ERA
Tristan LIM Ming Soon
Dr. Patrick THNG
MOBILE TRADING APPLICATION
Dr. Paul Robert GRIFFIN
AUTOMATED THEME SEARCH IN ICO WHITEPAPERS
Dr. Paul Robert GRIFFIN
ONLINE BANKING SECURITY PROBLEMS AND RISK MANAGEMENT
Mr. Edgar LOW
Click on the accordians to view a sample list of capstone projects undertaken by our MITB students, as well as the scope of work and learning objectives achieved.
Financial Technology & Analytics Track
Scope of Work: Explore Optical Character Recognition (OCR) technology to convert digital images containing texts, to machine readable and editable text formats which can be used for downstream data processing. It provides a methodology to automate data entry thereby reducing the processing time of a workflow considerably.
Implementation of OCR into the current workflow will automate indexing of data thereby providing the following benefits:
- Reduced processing time, as the data fields of the WFI system are automatically captured from the scanned image and populated.
- Straight through processing, reduced human intervention.
- Delivers accurate data and eliminates human processing errors.
- Scalable to large volumes and other work groups.
Learning Objectives Achieved: This project had exposed student to learn and understand the business problems that the banks encountered. The learning outcomes are as follows:
- Better understanding of OCR as a business solution.
- Designing different types of business documents such as RFP, Vendor Evaluation Matrix etc.
- Evaluate vendor products to address business needs.
- Re-engineer existing work flows to reap maximum benefits.
- Liaise with different stakeholders and vendors to come up with a solution for a business problem.
- Procurement of suitable vendors and products to address key business needs.
- Understanding the regulators’ guidelines for vendor procurement and abiding to the same.
- Understanding the process of procurement which is practised in leading banks.
Scope of Work: To understand existing processes in the Infrastructure Services area and then mapping it into a process flow diagram using Bizagi and Tibco process modeler. There will be deep dive analysis with the owners to come up with process improvement and optimization by analyzing each of the processes and explore what can be automated/outsourced/improvised so that time-to-market with a global and consistent quality is achieved. To create a document (pictorial and word) that will represent the process flow for a banking business unit and determine the business process challenges faced by the unit. The outcome of the project is to identify and recommend ways to overcome those challenges, re-engineer and optimize the current process flow. The scope of work covers the following:
- Discussion with process owners to gather their inputs on challenges faced.
- Study documents, interact with process owners to understand processes.
- Use Bizagi Process Modeler to draw the process flow diagram to illustrate understanding of the flow.
- Evaluate the processing time for the flow.
- Recommend changes, improvements and the new process flow.
Learning Objectives Achieved: This project had exposed the student to learn and understand the business process problems that the banks encountered. The learning outcomes are as follows:
- Overall understanding of each business process flow and performance.
- A detailed understanding of factors contributing the most time in the process flow.
- Manual tasks involved in performing the process.
- An understanding of the entire process flow.
- How a certain delay will impact the delivery time, thereby instilling a sense of responsibility.
- Recommended new process flow and re-engineered process flow.
- Student learnt how to re-engineer a process and how to plan effectively to meet deadlines.
- Student learnt how to manage the difficulties in getting information from business units.
Scope of Work: This project aims to increase the efficiency and the effectiveness of the AML process in a bank. The scope involves an analysis of the current AML process and ways to refine systems to improve their effectiveness. The project will articulate the introduction of the AML program, the current AML process, AML Reporting Automation system, and Portable AML system with the end objective of recommending improvements to the system. This project involves a detailed structure of the AML program, survey and system architecture and process.
Learning Objectives Achieved: This project had exposed student to learn and understand the AML system, challenges and ways to make the system more responsive and effective. The learning outcomes are as follows:
- The project enabled student to understand the investment and resources spent on AML system for the past decade, ways to optimize the process and reducing the manual work in order to reap the benefits of AML investment.
- Ways to deliver AML cases to the branch officers effectively so as to reduce the overdue and rejected rate of AML cases.
- Design of mobile application and a new architecture to connect the core AML system with the mobile application.
- Recommended improvement to the AML system to enhance the effectiveness of the AML process of banks.
Scope of Work: The main aim of this study is to deepen and update our understanding on the dynamics of political violence and its diffusion over spatial boundaries. Most studies of war in international relations, qualitative or otherwise adopt the assumption that states are the primary geopolitical actors in conflicts, with the country-dyad-year as the main unit of analysis, and war is often interpreted as the aggregation of violence or conflict events occurring between political entities. Explanations of the causes of war, conflict, and violence rely on the attributes and bilateral interstate relations of states as main independent variables with the goal of uncovering general patterns of behavior that are treated as universal tendencies. An update to this understanding of the role of space in conflict is needed, which will not only result in a more sophisticated appreciation of the effect of geography on conflict, but also alter the assumptions of how power is conceptualized and operationalized through spatial dimensions.
The scope of the project involved:
- Joining and cleaning political violence event data, country data, and geographical shapefiles
- Performing exploratory data analysis (EDA) on the combined dataset
- Testing various measures of spatial autocorrelation through exploratory spatial data analysis (ESDA) methods
- Visualising the data in an intuitive fashion for data analyses
- Validating findings through historical case studies.
Learning Objectives Achieved: This project has exposed the student to geospatial analytics and knowledge discovery methods, and provided an alternative understanding of how political violence spreads. The learning outcomes are as follows:
- Learning how to pre-process and visualise geospatial shape files and event data
- Learning how to conduct a proper and thorough exploratory data analysis and exploratory spatial data analysis in depth on large geospatial datasets across a time period
- Mastering data visualisation for time series and geospatial data
- Understanding how political violence spreads across geographical and national boundaries
- Performing the preprocessing and analysis from start to end in an R environment, including file manipulation
Reference Paper: Mack, V. Z., & Kam, T. S. (2018, November). Is There Space for Violence?: A Data-driven Approach to the Exploration of Spatial-Temporal Dimensions of Conflict. In Proceedings of the 2nd ACM SIGSPATIAL Workshop on Geospatial Humanities(p. 1). ACM.
Scope of Work: This study aims to support the data analysis needs of energy market using publicly available data and open source tools. The project will showcase the usage of R and various well-documented R packages in enabling an end-to-end analytics process. Besides performing analytics on the data, this project also aims to generate meaningful visualizations that enable stakeholders to understand the analysis results and use the output to aid them in decision making. The project first explores spatial autocorrelation of electricity consumption across different dwelling types in Singapore public housing. Then, the project goes on to estimate consumption amounts at various domain levels based on data provided at the postal code level.
Learning Objectives Achieved: This project had exposed student to learn about and understand the Singapore electricity market. The learning outcomes are as follows:
- Ability to work and produce quality output within a tight schedule (in response to a conference call for participation)
- Designed an end-to-end process using open-source tools to ingest, clean, prepare, analyse and visualize the data
- The importance of clear documentation in ensuring code readability and enabling easy code debugging and maintenance
- Problem solving skills and resourcefulness as the student learned to use a new package and had to use available online resources to overcome various errors
Artificial Intelligence Track
Scope of Work
Vehicle Routing Problem (VRP) is a combinatorial optimization and integer programming problem which is to find a set of optimal routes for a fleet of couriers to deliver a batch of job with minimum costs. In order to solve the VRP for optimum routes, two main approaches are purposed, namely the exact and heuristic algorithms. The exact algorithms provide optimal guaranteed solutions. However, they are computationally expensive due to high computational complexity. The heuristic algorithms are relatively fast. However, they are not guaranteed to generate the optimal solution. To solve large scale VRP, heuristic algorithms are commonly adopted considering the trade-off between optimality and computational cost. Suboptimal solutions could be developed be heuristic algorithms.
Motivated by the advancements in machine learning based studies, the project would develop an end-to-end model to learn heuristics directly from data and solve the vehicle routing problem.
- Develop the deep neural network model to solve vehicle routing problem with fast response time
- Develop the deep neural network model to solve vehicle routing problem with time window with fast response time
- Develop deep reinforcement learning framework to train the deep neural network for VRP and VRPTW without labelled data
Learning Objectives Achieved
The capstone project developed an attention based deep neural network model with encoder and decoder structure, which modeled the vehicle routing problem into a sequence problem similar to the natural language sequence. The attention model is able to solve the vehicle routing problem effectively, given a list of input unordered nodes, the model is able to produce efficient routes with fast response of 1ms/epoch.
The project further extend the vehicle routing problem into vehicle routing problem with time window by impose an extra time window constrain. By training the model with extra features of time window and delay penalty, the model is able to produce valid and efficient routes for vehicle routing problem with time window.
The project also developed a reinforcement learning framework based on REINFORCE algorithm by carefully design the reward functions and the environment functions. The project trained both the VRP model and VRPTW model without the need of labelled dataset.
Scope of Work
A Q&A forum has been developed for the undergraduate course – Spreadsheets Modelling and Analytics, which employs intelligent engagement features for collaborative learning. To further increase student engagement and active participation, the current system can be enhanced with auto-generation of tags to the posts and display of similar and related Q&A posts. The auto tags enable the students to search for the posts easily. Selecting auto-tags over manual tags provides the advantage of standardization and consistency across the system. Displaying similar and related posts helps the students to refer to the previous posts to derive an answer before the question is viewed and answered by other students.
The objectives of the capstone project are
- To automatically generate tags for a new post in the Q&A forum
- To retrieve similar posts from the Q&A system
- To retrieve related posts using the Knowledge Graph
Learning Objectives Achieved
Automatically generating a tag to a new post adds great value in improving the student engagement. It makes it easier for students to respond to questions with the topics that they are most familiar with, thus increasing the participation rate. The Labeled LDA, a supervised topic model is the most suited model for the given dataset, since it can effectively perform contextual word sense disambiguation since each post is associated with a labeled topic. The model also performs better than the baseline model of unsupervised LDA for multi-label classification. For extracting related posts, the usage of Knowledge Graph makes it easy to extract contextually similar posts when compared to the traditional relational database.
Interested company sponsors will be required to provide the following information which will be shared with the students:
- Brief description of the project
- Role of the student
- Skill sets required (technical and non-technical)
- Project duration
- Expected deliverables
The SMU MITB Office will liaise with the company to forward the profiles of interested students for consideration. The company selects the students via a process of interviews. Upon confirmation from the company, the SMU MITB Office will formalise the sponsorship and assign a suitable SMU faculty to supervise the project.
We welcome interested company sponsors to contact us at firstname.lastname@example.org.
Last updated on 14 Sep 2021 .