Data Science Portfolio

Here are some of my best Data Science Projects. I have explored various machine-learning algorithms for different datasets. Feel free to contanct me to learn more about my experience working with these projects.

What other factors besides the number of bedrooms can influence price of a house

Skills used: Python, pandas, sklearn, numpy, matplotlib, seaborn

Project Objective: Create a linear model using the provided data. The model needs to acquire an accurate final price prediction of a house using the provided features.

Quantifiable result: My model was able to acquire a validation score of 92%.

I used an XGBoost model to achieve the highest score.
A combinantion of GridSearchCV and my knowledge was used to tune my models.
I tested with multiple different models.
Used a pipeline to speed up my models.

Understanding how policies can be improved for employee benefits

Skills used: Python, sklearn, pandas, numpy, scipy, matplotlib, seaborn

Project Objective: Clean and organize the given data. Then, Cluster the data using a clustering algorithm and find a meaning or a message from the data.

Quantifiable result: I was able to get 5 solid clusters.

I used to the elbow method to certify the clusters.
During the EDA I was able to reduce a lot of noises.
I used data that was withing 2 standard deviation.
I used WCSS algorithm to determine the optimal number of clusters.

Examining the effect of environmental factors and weather on Bike rentals

Skills used: Python, pandas, numpy, matplotlib, plotly

Project Objective: Create a linear model and identify what features affects the number of bikes rented.

Quantifiable result: I was able to achieve a validation score of 74% using Bagging Model.

I used the the highest correlated Data.
I normalized my data in order to achieve a more accurate result.
The training score was a very high 96%.
My error margin was also calculated using the proper metrics.

How a target clients for a particular marketing campaign using many specific features

Skills used: Python, pandas, matplotlib, seaborn, sklearn, SMOTE

Project Objective: Using a vast data/features of bank customers, create a model that would determine which customers are likely to subscribe to a term deposit.

Quantifiable result: I was able to get a accuracy score of 90%

Logistic regression came out to be a very effective model.
I used a 0.30 test size and it gave me the best results.
I used a dummies to encode the categoriclas.
I used SMOTE because the two classes were highly imbalanced.

Image recognition using Convolutional Neural Network

Skills used: Python, tensorflow, keras, numpy

Project Objective: Create a model that could look at a picture, and specify if it is a cat or dog.

Quantifiable result: I was able to get a accuracy score of 90%.

I trained the data using 4000 images.
I used all the proper bulilding steps to build my CNN model.
I used a the ‘sequential’ model and it was very effective.
I used ‘adam’ as the optimizer and ‘binary_crossentropy’ for loss for the optimum result.