Final year project themes

Supervisor: Dr Daniel Stamate

(1) Expert system in the area of your choice (to be specified: e.g. medical)

Programmes: CS, CIS, IT

Project complexity and programme of study suitability: depends on the functionality of the system and language for coding it (e.g. CLIPS, FuzzyCLIPS, Prolog, Java, C++, C#, PHP+SQL, etc)

(2) Social Web Mining

Popular social networks such as Facebook and Twitter generate very large volumes of useful data on various topics. Who talk to whom, what they talk about, how often they talk? The social web mining has a much broader context, and its current development is based on increasing commercial benefits [1].

This project concerns the implementation of a software system capable to analyse data from social networks in order to find interesting patterns. The project requires Machine Learning and Natural Language Processing algorithm coding (e.g. clustering), using a programming language as Python.

Programmes: CS

Project complexity: depends on the chosen functionality of the system. Programming in Python.

(3) Automatic medical diagnosing

The project assumes the coding of a program that learns to diagnose patients. The program would analyse collected data regarding past diagnosed patients (namely characteristics as for instance their symptoms, test results, age, gender, etc, but also the result of the diagnosis - an ailment). The program learns from data about former diagnosed patients, by  automatically finding patterns or rules linking the patient characteristics and their diagnoses. These patterns/rules can be utilised for diagnosing new patients. The system can be used for information and educational purposes, or can be a useful complementary  tool that medical doctors can employ in diagnosing.  

The project requires Machine Learning / Data Mining algorithm coding (e.g. Decision Trees with C4.5 training algorithm, K- Nearest Neighbour, Neural Networks with Back Propagation or Genetic Learning training, Bayesian classifiers, possibly additional meta learning techniques, etc) using a programming language.

Programmes: CS, CIS

Project complexity and programme of study suitability: depends on the algorithm choice, and the language for coding the system (e.g. Java, C++, C#)

(4) Intelligent system for customer retention

This project regards the task of building an intelligent system that predicts which customers of a service provider (e.g. telephone/internet) are likely to leave current service for that of a competitor. These customers are called churners. The intelligent system would help retaining as many customers as possible by making offers (as for instance weekend free call time) to predicted churners.

The problem of customer churns is closely related to Customer Analytics.

The project requires Machine Learning / Data Mining algorithm coding (e.g. Decision Trees with C4.5 training algorithm, K- Nearest Neighbour, Neural Networks with Back Propagation or Genetic Learning training, Bayesian classifiers, possibly additional meta learning techniques, etc) using a programming language.

Programmes: CS, CIS 

Project complexity and programme of study suitability: depends on the algorithm choice, and the language for coding the system (e.g. Java, C++, C#)

(5) Intelligent system for detecting and profiling good and bad credit risk applicants

The project assumes the coding of an intelligent system that predicts which loan applicants are likely to be good or bad credit risk. This type of system is regularly used in banks to help deciding which loan applications to accept or reject.

The problem of detecting good/bad credit risk customers is related to Customer Analytics.

The project requires Machine Learning / Data Mining algorithm coding (e.g. Decision Trees with C4.5 training algorithm, K- Nearest Neighbour, Neural Networks with Back Propagation or Genetic Learning training, Bayesian classifiers, possibly additional meta learning techniques, etc) using a programming language.

Programmes: CS, CIS 

Project complexity and programme of study suitability: depends on the algorithm choice, and the language for coding the system (e.g. Java, C++, C#)

(6)  Intelligent E-commerce system with Data Mining engine for market basket analysis

Currently many websites that sell products use data mining to better service customers and improve sales. For example, when one chooses a particular book on Amazon’s website www.amazon.com, one is offered information about a few books that may interest: “customers who bought this book also bought books ...”. This is called association rule, and can be derived through data mining techniques related to association analysis. Particular algorithms can applied to data concerning previous baskets in order to derive the association rules efficiently. The project involves an implementation of a data mining engine – basically consisting in the implementation of the algorithm called Apriori on top of a database storing information about products and baskets. Obviously, in addition to the data mining engine, the project will include also a proper database design, and a web interface having functionalities as storing details about customers, storing the baskets, etc. 

Programmes: CS, CIS 

Project complexity and programme of study suitability: depends on the algorithm choice, and the languages for coding the system (e.g. Java, C++, C# for the data mining engine and SQL+PHP for the database and the web interface)