Choosing a thesis topic?

Question

Java

Hello everyone, I ask for help with advice on the topic of the question.
A student of the 5th year, specialty 010501 "Applied Mathematics and Informatics" ,
could not get the topic from the supervisor, he said to figure it out himself.
I've been struggling with this question for almost half a year and have not been able to find any interesting idea, and the deadlines are already running out. On the other hand, I do not want the diploma to be written "on the table".
My area of ​​interest and expertise is:

• java SE,
• familiar with Java EE,
• frontend technologies,
• data mining and machine learning (wrote term papers on reviews of clustering and classification algorithms, participated in solving competitive problems on Kaggle ),
• neural networks.

It so happened that there is no work experience in the specialty, so there is a desire that the diploma could be presented as a portfolio (I plan to develop as a Java programmer).
Another question is that the diploma requires the relevance of the problem, the use of a mathematical apparatus (therefore, options like a regular web portal are not considered) and the validity of the choice of programming language (it is clear that a tool is chosen for the task, but I get it differently), and therefore the range of tasks is somewhat limited.
There is a desire to work with Hadoop and / or Mahout, but again I do not know what this can be applied to.
PS There is another interest - forecasting sports events, in particular, I have been playing on the sports betting exchange for several years and even have some ideas for creating a project in this area, but I doubt its suitability as a thesis.

Answer 1

Good afternoon. I can offer a couple of options, but I immediately warn you - I am far from a great specialist in the topic that interests you.
Hadoop :
1. On Hadoop and Lucene, you can try to implement your own distributed full-text search engine that can index a huge array of data in a cluster.
2. You can try to implement the transformation of some SQL queries into tasks for the Hadoop cluster. For example, there is a very large amount of data stored in the cluster. You want to get some statistics on it. Write a query in the usual SQL form, it is transformed into tasks for Hadoop, executed, and you get the result.
Forecasting: instead of a sports betting, you can pay attention to real exchanges: currency, stock, etc. and write a robot that would implement a certain trading strategy. At 010501, similar topics for graduation theses were regularly taken from us.

Answer 2

Data-mining requires large amounts of information "at hand". Start from this when choosing a topic in the first place.
- search for botnet infections using the logs of a corporate proxy server
- anti-spam
- search for unusual activity of corporate network users
- identification of authorship (identification of an anonymous author) on Internet forums
- search for dependencies between news on information feeds and / or real events
- forecasting of any exchange rates based on news sites
- searching for dependencies between user actions in social networks or blogs
in general, fantasy with inputs like yours is virtually unlimited