Author : Ashay Vivek Sant and S. R. Khonde

As a business requirement, the advertiser codes need to be mapped to the full forms using Machine Learning Approach. This needs to be performed on production using a dedicated microservice. This will aid us in reducing the database lookup calls, and also be helpful for auto-suggestions for the user when creating new data. The data related to campaigns, the agreements, the flight time, date, amount, brand mapping is already present in a oracle based relational schema and is already normalized. This data needs to be combined and treated as a training data-set for learning a suitable model. The model learnt should also provide the functionality for auto-completion on the User Interface. The model learnt should be so generic that it could be tuned for any dataset in the near future, thereby making it potential candidate for prediction in advertisement world. For learning models an ensemble of classifiers could be run in parallel and then majority voting could be performed to decide the best classification. Simple Algorithms like that Naive Bayes, hyperplane learning algorithms like SVM and regression could be used along with unsupervised approaches like that of clustering.

Keywords: Machine, Ensemble Of Classifiers, Bayes Theorem, Support Vector Machines, SVM, Linear Regression, Lookup, Agency,           Sellers,   Clustering,           Unsupervised,      Supervised, Reinforcement