Data Science portfolio

Telco Churn Analytics

From headmap, as our target variable is [Churn], thus I focus on variable[churn]. It is postive correlated with variable [SeniorCitizen], [Partner], [InternetService_DSL], and [InternetService_Fiber optic], which mean the rate of churn is relatively depend on this variables. Look at variable [Contract], can conclude the longer contract of customer having subscribed, the rate of churn is low, and vice versa. Look at variable [tenute], can conclude the longer tenure of customer having subscribed, the rate of churn is low, and vice versa.

From the decision tree, we can conclude > Shorter tenure and MonthlyCharges > 686.5, the rate of churn is pretty high

Click here to view Colab notebook

Close Window

Regression Analysis - Social Ads Forecasting

Which social ads platform that you would invest in if you are given an allocation of RM100k advertising budget? This project gives insights on which social media platform generates higher revenue to the company.

Every thousand dollar we spend on facebook, it can bring more product sold. If we didn't allocate any budget on instagram, then still can more product sold. Which mean we can decide not need to spend money to advertise by instagram. Even though the R-square score for google is the highest among them, but look at the predicted revenue will be return when budget is 20k, surprisingly instagram is the highest among them

Click here to view Colab notebook

Close Window

NLTK - Analyze Sentiments for Amazon Alexa

Applying wordclouds and NLTK (Natural Language Toolkit), to gain insights on what are the most frequent words used each rating given and the sentiments of the reviews.

Using following keywords for test cases : Amazon Echo, Apple Music, sound quality, speaker connectivity, and google amazon. Look at the TF-IDF score for [Apple Music], [apple] is most heavier weight [0.9014958185243507] and very obvious that, [apple] is rarer than [music]. [sound quality] is the highest among other four keyword combination, and its similarity score : 0.85478 Review: "Good quality, great sound quality"

Click here to view Colab notebook

Close Window

Dashboard - HR Analytic

To understand the employee turnover in the organization, we use the Google Data Studio to present our findings

Click here to view Google Data Studio

Close Window

DS Project - Analyze sentiments and rating for Asian Tigers SG Customer review

I concluded the overall of the review based on sentiments analysis as following; Number of neutral of reviews: 593, which is 53.52% of the population. Number of positive of reviews: 504, which is 45.49% of the population. Number of negative of reviews: 12, which is 1.08% of the population. This is highest neutral population among other sentiment, one of the reason should be due to no review is filled in. Generally, this is quite good as the negative population pretty low. Which mean we still perform well in customer satisfaction.

Click here to view Colab notebook

Click here to view Customer Review Dashboard

Close Window