Dear All, I need a little help at starting my research work, I want to develop a system based on the complex event processing system to give the best offer to the customer, I mean the system should have ability to give customers best offer based on his behavior within 2 seconds? For example, if the user changes his address, in 2 second, the bank ubi (for example) will tell him that he can get a 20 % discount on purchases if he use his ubi credit card in this zone
My question is, what kind of data mining and machine learning algorithms which I have need to use it for starting my research work? please can anyone help me
Search for Machine Learning algorithms for recommendations.
You may start here:
https://medium.com/recombee-blog/machine-learning-for-recommender-systems-part-1-algorithms-evaluation-and-cold-start-6f696683d0ed
Youssef Idriss ,
You should break your question into two parts. The first one is about the algorithm used for ability to give customers best offer based on his behavior. To answer this question it is necessary to have at least an idea of the data being collected from the user to get the response(does it require text processing and mining? are there just numeric fields? if so are they just discrete or all continuous like your geo reference example? among other relevant pieces of information on data collection and processing).
The second one is on the performance constraint. This is not solely based on the algorithm, but further contemplate:
- The traffic that the application will have
- What resources are being used to scale up the service to get the response time desirted in terms of database(which also needs to contemplate the data and its relationship ), server configuration (cluster configuration, etc.), and network resources (router and switch capacity and configuration).
More information is needed to properly answer the question
That sounds like a rule engine: you have a body of rules, and the program must detect which apply.
The structure of a machine learning problem is that you have a training set of examples, each with an observable value and a label, and must learn an algorithm that assigns labels to future observables. You don't mention a training set. Without one, your problem is not machine learning.
Dear Youssef Idriss
To continue the point made by Chris Morris, ML means rule base but it should be CREATING not just maintaining the rules. In my view of data science, the one most important step to take is definition of intrinsic patterns (groups) of behavior. Once the groups are identified and verified, most of algorithms should do :)
How to verify group definitions is another big question. As above, my advice is to stick with easy to apply method. The complex ones may be heavy due to their generalization/ In the specific case under study a simple approach may be of no less of successful results.
Edith
- Badges
- Science topic