Dear Zafar,

I would suggest you to consider designing a Recommender System  that can help recommend potential candidates for political/non-political  positions/offices based on factors such as:

1. Candidate's current political achievements

2. Candidate's previous political achievments

3. Candidate's contributions to societal development

4. Candidate's educational background

5. Candidate's family background

6,  Candidate's financial status

6. Candidate's careerism, and

among several other attributes.

This kind of system would automatically rank potential candidates to be considered for a particular   office/position

I believe you can explore this idea and develop it further

I hope this information will give you some insight on new research direction in the field of Recommendation systems

Good luck.

Dear Oluwarotimi Williams Samuel,

I appreciate your suggestions and i hope that such kind of recommender systems have not been developed so far. However, i would like to get some more suggestions regarding its need and potential applications in real life. Moreover, what data set we can use for this problem?

Regards 

Zafar Ali

Dear Zafar,

I have the following suggestions:

1. Recommender system that will recommend books to different categories of student (Arts, Social Sciences, Sciences etc.) based on the examinations bodies syllabi such as WAEC, NECO, UTME, IJMB etc.

2. Career Counselling Recommender system that will recommend the combination of subjects that a student need to pass in order to study a particular course at polytechnics or universities.

I hope it will be of help.

Thanks

Dear Abba Almu,

I appreciate the ideas you suggested in your comments, these are very valuable suggestions for me. Can you please elaborate and suggest some factors for designing career counselling recommender system so that we can explore the area properly.

Regards

Zafar Ali

 I suggest you begin with reviewing literature on applications of recommender systems such as the following to figure out potential applications.

• http://web.stanford.edu/class/ee378b/papers/adomavicius-recsys.pdf
• http://arxiv.org/pdf/1006.5278v4.pdf

You can also explore the extension of existing systems in different languages and contexts. 

Sisay Chala,

Thanks for your kind suggestions, i will read these articles. Actually i am trying to find some new areas of research where recommender systems have not been introduced so far. As Abba Almu mentioned that it such systems can be used in domains of career Counselling and books as well.

Regards

Zafar Ali

A recommender system can as well be useful in researcher to researcher matchmaking for collaborations based on their areas of interest.

Chinedu Pascal Ezenkwu,

The idea you presented is of great value, however what data set we can use for this purpose to classify researchers. i mean to say that if a researcher has to be classified in our system then what kind of parameters we can use for that purpose and which collaborative approach will yield better results.?

Just like researchgate is doing to recommend you questions related to your expertise, you may analyse publications and/or profile data to do the matching.

Sisay Chala,

But, if a researcher don't have publication in that case how can we classify and assign him/her to a class of researchers? How we can fix this problem?

To solve a Cold-start researcher problem - community detection algorithm or clustering algorithm should be able to classify the researcher by following one of these direction :

(1)  User with absolutely zero information (i.e., no rating, no publication, no digital trail at all) - random classifier should do. Since their interests change over time, a real-time classifier is required;

(2) User with some trails -

• Classify them by their user profiles : publications based on keywords, titles, journal title/conf. proc. title;
• Based on their clickstream data.

Dear Oluwarotimi Williams Samuel,

Is there any other information and potential applications of such system you have recommended, i mean what could be the application and is it of great importance to recommend a politician? 

Abdul Hadi Mohamad,

I appreciate your proposed approach to resolve cold start problem, but the question is that on which data we should train our system, i mean we should use research articles to extract features or questions for that purpose. How our system will initially start working and become functional.