Every ML project is unique in its own ways. Although, for each such project, some predefined steps can be followed.
There is no such strict flow engraved in stone, but a general template can be proposed.
Prepare the problem
You first need to understand the situation and the problem which needs to be solved and metrics required.
Load & Summarize the Data
After the data is loaded, you need to check how the data looks and what all it contains.
Initial steps include descriptive statistics, correlations and other visualizations to summarize the features and target.
Prepare the Data
Getting data into the required format and pre processing it by impuation, scaling.
Feature engineering is done on the basis of feature importances and impact to remove redundancy and reduce constraints.
Evaluate Algorithms
Once the data is ready, you’d need to check the performance of various algorithms.
The data is split into multiple folds and then trained & validated on them, also called as cross validation.
Improve Accuracy
Once you have the best algorithms, you’d need to improve them by tuning and making ensembles.
Finalize the model
The best model is saved and deployed for use, and later retrained.
#datascience #machinelearning