This project aims to perform search string optimization in the search process of Ebay consumer website. Access to the Ebay Api has been legally obtained through public developer keys.

Objective

Given a search string, query the complete ebay database and list the top 100 results which match the context and topics of the passed search string.

This projects uses standard NLTK python distribution and NLP algorithms to find a consistent measure of similarity and tries to match inventory items along with the search string. It must be noted that the comparison process must be fast and efficient in nature so as to be able to generate a working, deployable system. Once the results are finalized, the list is displayed and that would be the list displayed in the search results when actually implemented.

This project aims to solve Quoras question similarity challenge. The data used is the publicly released data by Quora regarding training and testing material for the challenge.

Objective

Given a user attempting to submit a question, check the similarity of that question with already existing questions and suggest if it is a duplicate or not.

This projects uses standard NLTK python distribution and NLP algorithms to find a consistent measure of similarity and tries to match the current question along with the existing questions. A match threshold can be implemented and if there is a similarity beyond that, then it can be deemed as duplicate.

This project aims to predict scores and mood of students who are the main sources of surveys. It is a way to understand the non verbal aspects gleaned by students within the tutoring atmosphere.

Objective

Given a lot of surveys, estimate the mood of the tutoring session. Alo, estimate the missing scores in the survey.

LAC is a tutoring service provided to bachelors students at the University of Cincinnati. There are 5 major departments and each deal with a unique tutoring atmosphere. Each student who undergoes these kind of sessions are generally required to submit surveys regarding their experiences. These surveys help the decision makers know the performance of a session and its tutor. This project aims to automate the process of estimating the mood. Different kinds of survey text and words are aggregated into dictionaries representing level of mood and performance expected from a session. Then NLP is used to remove noisy parts within a prospective survey and make a similarity/context match with a pre-existing expected level. Also, in case a score is missing, the mood is utilized to populate the score.

This project is a concrete example of a full life cycle data science project. Data is collected, cleaned, analyzed and conclusions are made with a well developed website kind of visualization.

Objective

Given Airbnb rental data and Zillow housing data, estimate the best zipcodes in New York city to invest so as to have best results.

A zipcode which provides a decent amount of revenue, backed by a good house market value and the prospect of higher business opportunity is supposed to be a good candidate and should be in the chosen list of zipcodes to invest. Thus, such individual aspects need to be analyzed leading to a conclusion regarding the final candidate list of zipcodes. There are several factors such as Revenue Analysis, Return on Investment, Market Concentration, Real estate market values and tourism prospects etc. which are analyzed before arriving at a sane conclusion.

Check out the code and download the documentation module to see the results for yourself !!

This project is a concrete example of a full life cycle data science project. Data is collected, cleaned, analyzed and conclusions are made with different kind of insightful visualizations.

Objective

Given House loan market data, analyze the eastern states of USA and find which state is better for a new entrant and according to what conditions.

There are several critical factors like population, rate of interest, future forecast etc. to consider before entering the business which are covered to support the analysis. All such parameters have been analysed and the visuals for the same can be generated using the code in the github link.

This project is a concrete example of a full life cycle data visualization project. Data is collected, cleaned, analyzed and conclusions are made with different kind of insightful visualizations.

Objective

Given, Flying-Etiquette survey data, perform visualizations from various aspects and angles to bring out concrete insights.

There are various fields which provide information about the view of people using flight as the medium of transport. Although each field itself provides a good insight on its own, two or more fields can be compared amongst each other to generate more comparative insights. The same has been done in this project and many interesting visualizations have been crafted out. The whole presentation has been arranged in the form of a website and it is present inside the repository link below. Instructions to access the same are present in the Readme file.

This project is a full life cycle data science work. Data is collected, cleaned, modelled and researched, simulated with optimizations and finally followed by insightful visualizations.

Objective

This project shows how the performance of the basic Local Linear Wavelet Neural Network model (LLWNN) can be improved with hybridizing it with a fuzzy model. The new improved LLWNN based Neurofuzzy hybrid model is used to predict two currency exchange rates i. e. the U. S. Dollar to the Indian Rupee and the U. S. Dollar to the Japanese Yen. The forecasting of foreign exchange rates is done on different time horizons for 1 day, 1 week and 1 month ahead. The LLWNN and Neurofuzzy hybrid models are trained with the backpropagation training algorithm. The two performance measurers i. e. the Root Mean Square Error (RMSE) and the Mean Absolute Percentage Error (MAPE) show the superiority of the Neurofuzzy hybrid model over the LLWNN model.

This project is a full life cycle data science work. Data is collected, cleaned, modelled and researched, simulated with optimizations and finally followed by insightful visualizations.

Objective

Given stock market data from major stock sources, predict the volatility and changes of the values. Also, optimize the algorithms with swarm inspired methods.

Swarm networks are a new class of neural networks which are inspired by swarm intelligence. Swarm intelligence is the property of co-ordinated behaviour seen amongst the social organisms. The simple local interaction of the swarm members results in complex and intelligent global behaviour. This phenomenon is adapted in swarm intelligence so as to solve many problems.This paper presents a comparative study of particle swarm optimization (PSO) based hybrid swarmnet and simple FLANN model. Here both the models are initially trained with LMS algorithm, then with PSO algorithm. The models are forecasting the stock indices of two different datasets i.e. NIFTY and NASDAQ on different time horizons i.e. one day, one week, and one month ahead. The performance is evaluated on the basis of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). It was verified that PSO based hybrid swarmnet performed better in comparison to PSO based FLANN model, simple hybrid model trained with LMS and simple FLANN model trained with LMS.

This project is a full life cycle data science work. It is the topic of my Masters Thesis Research. Data is collected, cleaned, modelled and researched, simulated with optimizations and finally followed by insightful visualizations. Two innovative algorithms are research and modelled into achieving the segmentation operations.

Objective

Research innovative algorithms to perform the image segmentation operation in the process of skin cancer detection.

The purpose of this project is to research innovative segmentation algorithms that will be the part of skin cancer detection process. As a part of the thesis, two application specific modeled algorithms have been designed to perform the process of segmentation, which is the second step in the overall process of classification of the image into various cancerous categories. A novel attempt to use a clustering based algorithm to address a segmentation task has been attempted and achieved through this research. Images have been considered in the gray scale mode and an attempt has been made to extract maximum results without color information. Both algorithms developed involve training and testing phases. Also, they are inspired by the power of Neural Networks. Once the segmentation is done, various performance metrics have been calculated and reported along with visual aid regarding how well the segmentation occurred. The performance has also been compared with the commonly used methods in image segmentation and the advantages as well as performance factors are well critiqued and documented to provide a holistic view related to the usage of such algorithms in the concerned topic of skin cancer segmentation. Experimental testing has also been done with images having pre-known ground truth information and the resulting segmented portions as well as quality has been shown.