The goal of this project is to build a machine learning model that can predict whether a client will subscribe to a term deposit or not. This will predict which clients will likely subscribe the term deposit before agents make phone calls.It will also help the banks to decide how many clients needs to be contacted in order to meet their business target,based on which banks can plan the scope,budget and resources of marketing campaign accordingly. The dataset used in this project is obtained from kaggle.
Here is the link to the dataset: https://www.kaggle.com/janiobachmann/bank-marketing-dataset
The dataset includes details of about 11162 customers.
The features used in the data are as follows:
- age (numeric)
- job : type of job (categorical: 'admin.','blue-collar','entrepreneur','housemaid','management','retired','self-employed','services','student','technician','unemployed','unknown')
- marital : marital status (categorical: 'divorced','married','single','unknown'; note: 'divorced' means divorced or widowed)
- education (categorical: 'basic.4y','basic.6y','basic.9y','high.school','illiterate','professional.course','university.degree','unknown')
- default: has credit in default? (categorical: 'no','yes','unknown')
- housing: has housing loan? (categorical: 'no','yes','unknown')
- loan: has personal loan? (categorical: 'no','yes','unknown')
- contact: contact communication type (categorical: 'cellular','telephone')
- month: last contact month of year (categorical: 'jan', 'feb', 'mar', ..., 'nov', 'dec')
- day_of_week: last contact day of the week (categorical: 'mon','tue','wed','thu','fri')
- duration: last contact duration, in seconds (numeric). Important note: this attribute highly affects the output target (e.g., if duration=0 then y='no').
- campaign: number of contacts performed during this campaign and for this client (numeric, includes last contact)
- pdays: number of days that passed by after the client was last contacted from a previous campaign (numeric; 999 means client was not previously contacted)
- previous: number of contacts performed before this campaign and for this client (numeric)
- poutcome: outcome of the previous marketing campaign (categorical: 'failure','nonexistent','success') Output variable (desired target):
- has the client subscribed a term deposit? (binary: 'yes','no')
Exploring categorical features
From the above plots we can conclude that:
- In the dataset most of the clients have management as their job type.
- Clients who are married have more number of records in the given dataset.
- Most of the clients have secondary education.
- Default does not seems to be an important feature so it can be dropped.
From the above plots we can conclude that:
- Features like age and day are distributed normally.
- Features like balance,duration,campaign,pdays and previous are highly skewed towards left and seems to have some outliers.
It can be seen that age,balance,duration,pdays,previous have some outliers.
It can be seen that no feature is heavily correlated with any other feature.
Random forest classifier gives 0.85 accuracy score and 0.85 precision score Confusion matrix:
SVM classifier gives 0.72 accuracy score and 0.74 precision score.
Confusion matrix:
XGBoost classifier gives 0.85 accuracy score and 0.85 precision score
Confusion matrix:
A web app of this model is also created using flask and it is deployed on heroku.
You can view the web application here.
Screenshots of project:
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent