This page will be updated with my research as I progress through my PhD.<\/p>\n\n\n\n
PhD Mobile phone networks have strict terms and conditions which outline what customers can use their network for. However, a minority of users misuse their SIM cards for example by sending large numbers of text messages (spamming), having a high volume of phone calls or making a lot of contact with international or premium numbers. Often spam messages are associated with scams and these cost consumers millions of pounds each year. For this reason, and to protect the reputation of mobile network companies, we wish to develop methods to quickly identify spamming on the network so that the mobile network provider can take the necessary actions to protect customers.<\/p>\n This project is in partnership with Tesco Mobile to develop new methods of showing this network misuse by looking for anomalous patterns in their data. While some network misuse may be obvious, such malicious users are becoming more aware of simple detection methods and so more sophisticated ones are needed. These new anomaly detection methods aim to detect unusual patterns both within and between customer usage profiles. As part of this, we also want to characterise normal network usage to capture different typical usage profiles. Initial ideas for characterising this include using regime switching models or Hidden Markov models to model when users are in conversation or out of conversation with other users. Alternatively, their overall usage patterns may be clustered using features such as frequency of messages and number of recipients to identify similar typical usage patterns. Abnormal behaviour could then be identified by its large difference from existing clusters.\u00a0<\/p>\n\n\n MRes <\/span><\/strong> Research Topic 1 <\/span> Download RT1 Report<\/a> Research Topic 2<\/span>
<\/span><\/strong>My PhD subject is in developing new anomaly detection methods which can be used specifically by mobile phone companies to identify network misuse.\u00a0<\/p>\n
\n\n\n\n
During our MRes, we chose to research different topics and produce reports and presentations on them.<\/p>\n\n\n\n
We had to choose a topic from statistics or OR, and produce a popular science style report and presentation on the topic. I chose to do mine on Inventory Management methods, specifically for hospitals ordering blood units for their blood banks. The age-dependent periodic review model was developed in relation to stocking perishable goods in retail, however I tried to adapt this model to this hospital scenario to see how it compared to more traditional models such as the Newsvendor model.<\/p>\n\n\n\n
Download RT1 Presentation<\/a><\/p>\n\n\n\n
For my statistics topic, I chose N-mixture models within statistical ecology. These models are used to estimate the true population sizes from different survey sites when different numbers of surveys have been conducted. I was able to test these methods with data collected from fieldwork in Madagascar.\u00a0<\/p>\n