Data leads to insights, insight leads to products. Whether the product is about creating better customer segmentation, or developing superior product recommendation system, product makes some difference for your business. Within machine learning region, there are some "typical" problems to solve, mostly whether a training data has label or not. How to built up this model is a key part in machine learning.
While building model is challenging, explaining a model's result is also very important. I have seen many company keep using linear regression over other more advanced model for the solo purpose of "easily explainable" to their end-customer. If there could be an advanced model + good prediction explanation/illustration, then it would definitely be impressive.
Two packages are widely used in building models in Python: scikit-learn and statsmodels.
While building model is challenging, explaining a model's result is also very important. I have seen many company keep using linear regression over other more advanced model for the solo purpose of "easily explainable" to their end-customer. If there could be an advanced model + good prediction explanation/illustration, then it would definitely be impressive.
- Building model
- Supervised learning (regression, classification)
- Unsupervised learning (dimension reduction, clustering, outlier detection)
- Explaining model
- Global approach
- Local approach
Two packages are widely used in building models in Python: scikit-learn and statsmodels.
