Viewer Propensity Model

Database Name: Advertising.csv

Github Link: https://github.com/kkemwal1998/Market_Research

Nutshell: The project consists of the nutshell of the model, which has been constructed using Visio.

Database Source: https://www.kaggle.com/code/mafrojaakter/customer-ad-click-prediction

Database Description: The database consists of details about the number of people who clicked on ads or not, which has been signified in the form of binary classified language 0 and 1. (0 = Not click & 1= Click)

Project Objective: The project aims to build a predictive model that aims to forecast the probability of whether people will click on ads or not.

Impact: There are 90% true cases being predicted under the model.

Process: The model follows 5 steps:-

a) Database Acquisition: After the data was extracted from Kaggle in a CSV format, it was uploaded to Jupyter Notebook using the Pandas library.

b) EDA (Exploratory Data Analysis): The project consists of data visualization using Matplot and Seaborn Library. The graphs are:-

1) Histogram: Frequency distribution of age among the users.

2) Jointplot:

i) Income V/S Age

ii) Daily Spent V/S Age (Hue= Clicked on Ad)

iii) Daily time spent V/S Daily internet usage

3) Pairplot

c) Database Training & Testing:

i) Testing Dataset: 30% of the database.
ii) Training Dataset: 70% of the database.

d) Model Testing:
i) Classification Report:

- Precision score: 90

- Recall Score: 90

ii) Confusion Matrix:

- Accuracy = (True Negative + True Positive)/Total cases
= 140 + 129 = 269/300 = 90%

e) Model Deployment:

-The model has been deployed on the testing dataset with
the capability of predicting the outcome with an accuracy
of 90%.

-Using Sigmoid Activation function, the model's accuracy
has been increased to 97.74%.

-The prediction on the Model shows that the chances for the
user to click on ads are higher.