Even though deep learning is powerful and more accurate in predicting but its a black box ML. Can you imagine trusting a financial application where people down know how a part of it works. May be white box machine learning is all that can be used on financial data and the learning can be used to help business analyst and developers.
With this rationale, I started reading and experimenting with Decision Tree algorithms. I found a good place to start, “Tree-based Machine Learning Algorithms: Decision Trees, Random Forests, and Boosting” by Clinton Sheppard.
The book started with a simple algorithm, showed its limitations and then changed it to overcome the limitation. This helps to build a good understanding of the algorithm. I just finished the first 2 chapters. So far what I learnt can be applied on flat data, like a data from a single table. I need to see how to make the algorithm work on non-flat data (data from 2 tables joined by unique key and having 0..* relation.