Categorical Embedding

• Entity Embeddings
• example: Play store’s recommendation system

Beyond Deeplearning

• Ensembles of Decision trees (structured data)
• Multilayered NN’s (Unstructured data)

Analyzing Tabular data

Interpreting a model of tabular data is significantly easier for decision tree ensembles, for example, we can question— which row can do this or which cols has the most effect on the result

So decision trees are our first approach for analyzing tabular dataset.

There are exceptions:

• if there is high cardinality in the dataset, This refers to the number of discrete levels representing categories, something like zip code, which can take thousands of possible levels.
• other is cols that contain data that would be best understood with a neural network, example plain text.

Scikit learn is here to help!

This is a popular library for creating ML models using approaches that are NOT covered by Deep Learning.

Mean squared error and Root mean squared error

TODO

Decision trees

Think about this: when creating a decision tree, we ask a series of questions to split the data. How can we be sure that the questions we choose in this process are the right ones?

This is a greedy approach to choose the splitting question that produces the best split, i.e, that most accurately separates the items into two distinct categories and then apply the same rule to the group that split produces and so on.