Here you can check all of my pet projects which allowed me to experiment with the knowledge acquired though various MOOCs and books.I have updated some of them here.
Walton Analytics - Product Performance and Consumer Behavior Dashboard
- Merged data from different possible sources, cleaned the data and enriched it using Tableau Prep Builder
- Performed cohort analysis on the sales data to track the movement and flow of customers between differenmt quarters
- Perfomed RFM analysis to segment the customers into 5 different categories based on the managerial comfort and explored their overall profile
- Computed Customer Lifetime Value of all the customers and projevcted it to next 6 years based on the consumer behavior and their churn rate
TOOLS USED: Tableau Desktop, Tableau Prep Builder, RStudio
Embracing Subcultures - The success of Burger King
- Performed a case study on the series of marketing campaigns by Burger King over years embracing the subcultures existing within their customer profiles
- Researched the different kinds of subcultures they have utilized and the response among their customers
- Developed different methodologies on how they would have utilized the power of big data to perform these tasks
TOOLS USED: Google Sheets, Google Slides
Baker’s Database Analysis for HR
- Developed SQL queries for the requirements by HR from an employee database of 7 different tables and arouns 3 million entries
- Derived insights from the extracted data using Tableau Desktop to provide recommendations to the HR
- Provided recommendations to improve employee retainment in the company
TOOLS USED: MySQL Workbench, Tableau Desktop
CDiscount - Data Vizualization
- Cleaned and enriched the dataset using Tableau Prep Builder
- Compared the sales trend of CDiscount in France with other countries
- Explored the data, found the profit generating products and loss creating products
- Followed almost all the data visualization principles to obtain clutter free, clear data viz
TOOLS USED: Tableau Desktop, Tableau Prep Builder
Quora Duplicate Question prediction
- Repeated questions are one of the greatest complications of Quora. It leads to confusion and poor reach
- I processed the data using TF-IDF vectorizer and fit the model on various classification models
- Conventional logistic regression seem to give better solutions
TOOLS USED: Sklearn, pandas, numpy
Whatsapp Frequent Chat analysis
- Developed a model which can analyze whatsapp chats and deliver us valuable insights
- Determined the frequently used words, most active days and most active time of the day
- Working on determining the sentiment of chat and predicting the relationship based on the chat
TOOLS USED: Sklearn, pandas, numpy, matplotlib
Heart disease prediction
- Trained a ML model to predict the possibility of occurance of heart disease for a person based on his demographic attributes
- Some of the features used include cholestrol level, age, weight, gender, heart beat rate etc. and achieved 85% accuracy on the validation set
TOOLS USED: sklearn, numpy, pandas, seaborn
Tourism data analyis
- Performed data analysis on the world tourism data using SAS to uncover some hidden facts
- Found out the countries with maximum share of tourism attraction, country which generates maximum revenue, country which spends the maximum on tourism industry
- TOOLS USED: SAS
Malaria detection from microscope images
- Trained and improved the performance of a CNN architecture using various image augmentation techniques and optimizers
- Achieved an training and validation accuracy of both 94.5%
TOOLS USED: Keras, pickle, numpy, matplotlib
Digit recognizer web application
- Trained a custom developed CNN architecture to train on MNIST dataset
- The model is then served using flask to predict the digits on local server
TOOLS USED: Keras, pickle, numpy
Stackoverflow pragramming language classification
- Used questions and its respective programming lanuages to train a machine learning model with TF-IDF vectorizer
- Achieved 92% accuracy on predicting the programming langauge given its questions.
TOOLS USED: sklearn, scipy, NLTK, pandas
Cat vs no-cat classifier
- Trained a CNN model to predict between cats and non-cats images
- The model achieved a 91% accuracy in identifying
TOOLS USED: Keras, pickle, numpy
Text exploration using Topic Modelling
- Performed Topic modelling on a set of documents using LSA, LDA and NMF with available packages
- Implemented topic modelling papeers such as PAM, pLSA in python
TOOLS USED: sklearn, pandas, numpy, scipy
Indian Restaurant Market Analysis
- Preprocessd and feature engineered the data on Indian Restaurants spread across various cities
- Fit a model to determine the overall rating based on the most relevant features
TOOLS USED: dplyr, ggplot2
Customer Churn Prediction
- Exploed the data to detremine the significant features among the given features
- Model is fit using classification methods and the results are compared
TOOLS USED: Sklearn, pandas, numpy, matplotlib, seaborn
Design of Front and Rear Suspension Geometry for All-terrain vehicles
- Unequal Unparallel double wishbone system and H arm with camber link suspension system has been for Front and Rear Suspension Geometry respectively based on the design considerations
- The system is designed using Solidworks and is analysed using ANSYS and MATLAB
TOOLS USED: Solidworks, ANSYS, Lotus Shark
Development of Part Identification System of Baja vehicle components
- In a single baja vehicle more than 500 components are being used. Unnecessary time is used for searching different components during assembly
- So, various components are named using Hybrid Part Identification systems and are grouped together based on the frequency of usage
- This decreased the time consumption by many folds
TOOLS USED: MS Excel
Optimisation of Front and Rear Uprights for All-terrain vehicles
- Based on the load applied and fatigue life, structural design and sectional thickness has been modified to minimise the material usage without trading off the steering performance of the vehicle
- Final design is chosen based upon a number of iterations results and in accordance with suspension parameters too
TOOLS USED: Solidworks, ANSYS
Intelligent Anti-theft System for Two Wheelers
- Incase, if a two wheeler is attempted or subjected to theft, our installed set-up captures the snap of the intruder and sends the snap to the owner’s phone along with the dynamic GPS coordinates of the two wheeler using GSM
TOOLS USED: Arduino UNO R3
Design and Analysis of Hydrogen Propelled Mobility Vehicle
- A parallel hybrid two wheeler powered by both Li polymer battery with a rated capacity of 20Ah (based on our requirement) and PEMFC has been modeled with custom designed chassis, suspension, brakes and storage tank.
- Material for each component has been chosen from numerous analysis of the components’ solid model.
- Pedal force, brake force distribution, stopping distance and brake power are calculated theoretically.
- With a fixed wheel base of 1.32m and load distribution calculated based on brake power, all the suspension parameters such as motion ratio, corner weight, wheel rate, spring rate are calculated.
- These parameters are optimized using MS Excel and Matlab Simulink.
- The chassis body dimensions and cross sections are chosen based on Taguchi method.
- Real effort has been taken to measure the real dimensions of the chassis based on the ergonomic comfort of many persons of distributed age and size.
TOOLS USED: Soliworks, ANSYS, Minitab, Matlab Simulink, MS Excel
Design and Analysis of on-board high pressure compressed hydrogen storage tank
- After finalizing the hybrid vehicle architecture and storage pressure of 700 bar, power required to propel a payload of 180 kg was calculated.
- Comparing the heat of combustion of hydrogen and the energy needed to be produced, the mass of hydrogen required is calculated to be 1.2 kg.
- Volume of hydrogen required is calculated and compared by Ideal gas equation, van Der Wall’s equation and generalized comparability chart.
- A triple composite layer tank is chosen with specific dimensions after numerous iterations on ANSYS.
- Finally, analytical simulation of burst test, crush test is carried out on the final model to ensure its endurance
TOOLS USED: Soliworks, ANSYS, Matlab Simulink, MS Excel