Making A Model Is Like Baking A Cake

The Types of Cakes Available

As we progress further in our modern era, the advances of Data Science and Technology continue to make marvel strides in various fields of study and practice. As a result of the vast applicability of Data Science and Technology, various different types of models have been constructed.
To name a few: Generalized Linear Models, Support-Vector Machines, K-Nearest Neighbor Algorithm, Gradient Boosting Decision Trees, Random Forest, and Neural Networks.
Given the volume of the data and the complexity of the interactions within the data, various Data Science specific packages have been developed in a few different languages. Within python, we have, not exclusively, sklearn, xgboost, lightgbm, pyspark and H2O. Within R, we have, not exclusively, Caret, Prophet, SparkR, and xgboost.

Each of the aforementioned packages attempts to solve very specific Data Science problems. However, given the nature of the initial problem, we may need a more custom approach deriving the solution. Therein lies the problem each Data Scientist strives to solve.

Typically, due to the variety of the problems in the field, Data Scientists develops solutions using models, with predefined model architecture, and seeks to increase the performance of certain KPI metric performance the predefined model was not specifically made to solve.

Fortunately, there is a way to develop a modelling architecture to specifically solve your problem. By understanding the mathematical and statistical processes each of these predefined model architecture uses, it is possible to reverse engineer the model to specifically your problem.
First, let us examine one family of models, and then proceed to developing our own model....

To read understand and read more regarding how to create your own special unique ML custom model, please continue reading here: https://towardsdatascience.com/making-a-model-is-like-baking-a-cake-5f2443894c5f