Prediction-with-KNN-Classification

Implementation of KNN algorithm in Python 3 for prediction of Target class.

Description

• K-Nearest-Neighbors algorithm is used for classification and regression both.
• In this project, it is used for classification.
• Classified dataset used to predict target class.

Data set format

• CSV (Comma Separated Values) format.
• Attributes can be integer or real values.
• Responses can be integer, real or categorical.

Overview

The primary goal is to predict target class based on multiple independent variable.

liabrary

• pandas, numpy, matplotlib,seaborn,sklearn,joblib used in project

Methodology

1. Machine learning life cycle:

   • followed indistry standard practice of machine learning life cycle steps.

2. Preprocessing and EDA:

   • implement necessary transformation, preprocessing of dataset.
   • conduct exploratory data analysis on dataset.

3. Visualization:

   • visualised data using visualisation library like matplotlib, seaborn.

4. Algorithm:

   • scikit library use for KNN algorithm.

5. model validation:

   • model validate with accuracy score of diff K, confusion metrix.

6. save model:

   • joblib library used to dump model.
   • model is saved in .ipynb formate as i_phone_purchase_product_using_KNN_model.

EDA is not done in project, due to focus on machine learning only.

KNN model:

• K value by standard method is 31.
• K value by error method is 18 , so we considered 17 or 19 as odd number.
• model build with all k values and checked accuracy score and confusion metrix.
• with K=31, accuracy score is 94.5%
• with k=17, accuracy score is 95.5%
• with k= 19, accuracy score is 96.5%
• so accuracy in k=19 is better, so this model is saved and loaded to predict.