Project Walkthrough.txt

Yep, This project helps us to predict whether an employee should get promotion or not, based on the Train and Test Datasets collected from kaggle datastore.
I have imported some python packages, Machine learning Libraries, used a Machine Learning Algorithm called Decision Tree Classifier, in order to predict the Employee promotion status.

Steps Performed are:
1. Importing Datsets.(.csv files)

2. Reading Datasets(pd.read_csv('file_name'))

3. visualising the data inorder to know the relationship of the columns.
(using visualization tools such as piecharts,stripplot,countplot,bar,boxenplot,box,matplotlib,seaborn library etc)

4.OUTLIER DETECTION.
Identifying OUTLIERS in our datasets using boxplots.

5.Treatment of MISSING values.

6. Univariate Analysis(Analysing single column).

7. BiVariate Analysis(Analysing 2 columns and knowing their relationship)

(Categorical-Categorical, Categorical-Numerical, Numerical-Numerical)

8.Making Interactive Functionality using @interact_manual from ipywidgets library.

9.MULTI VARIATE Analysis, making correlation heatmaps(knowing which two columns have highest correlation and which doesn't)

10.Feature Engineering: Technique of identifying useful features from our datasets.

11. Grouping and Filtering: using groupby() and filter fucntions investigating our datasets.
Crosstab()-> helps us to identify how 1 column effects the other column.

12.Making @interact_manual between the columns.
(Categorical-Categorical, Categorical-Numerical, Numerical-Numerical)

13.Dealing With Categorical Columns(converting String-Numerical using Labelencoder and by traditional .replace())

14.Splitting our datasets(A=target_variable,B=rest of data)

15.Resampling: Process of creating sample of data from existing dataset.
using SMOTE algorithm which is an over_sampling method works to reduce imbalance in the dataset.
Techniques of Resampling: oversampling,undersampling,cluster-based_sampling

16.Feature Scaling: method of making all the features in our dataset share the same scale, if not used the higher values of 1 feature might dominate other values.
Techniques: Standardization range[0 mean and variance 1], Normalization range[0-1]

17.Predictive Modelling.
-> Decision Tree Classifier.
-> Train the model using resampled datasets.
-> Analyse the Training and Testing Accuracies, make a CONFUSION MATRIX and visualize them using heatmap and identify the errors made by the model.

18. Feature Selection of Decision Tree Model.
using RFECV() -> selects the best features/columns from the dateset and eliminates the features/columns which are not contributing to best prediction outcome.(it is all done recursively)

19.Prepare a Classification Report of (y_valid,y_predict) using the module classification_report imported from sklearn package.

20. Finally, check the descriptive stats for the dataset and make some Real Time Predictions.
	prediction = rfecv.predict.np_array(([[7,2,0,1,35,5,3,1,0,50,6,80,7,700]]))
	print("Whether the Employee should get a Promotion : 1-> Promotion, and 0-> No Promotion :", predictions)
	Whether the Employee should get a Promotion : 1-> Promotion, and 0-> No Promotion : [1]