Abhishek Singh

· Performed multivariate statistical analysis to determine importance of variables and its credibility.
· PCA was used to reduce dimensionality, feature engineering was used to extract more meaningful variables.
· A Classifier model was developed using Logistic Regression to predict 85% of the defaulters for the next month.

Performed data analysis on a credit scoring dataset and developed model to predict good and bad credit risk for future applicants based on the credit of past applicants.
• The Germen Credit data has 1000 past credit applicants, described by 30 predictors.
• Explored the data to identify the patterns and variables of interest in the dataset.
• Partitioned the data into training and validation and developed decision tree.
• Evaluated other types of decision tree – CART and J48 with… Show more

Performed data analysis on a credit scoring dataset and developed model to predict good and bad credit risk for future applicants based on the credit of past applicants.
• The Germen Credit data has 1000 past credit applicants, described by 30 predictors.
• Explored the data to identify the patterns and variables of interest in the dataset.
• Partitioned the data into training and validation and developed decision tree.
• Evaluated other types of decision tree – CART and J48 with different tree parameters and pruning.
• Calculated net profit and misclassification costs using the best model to determine the cutoff for the performance of future applicants.

Tools Used: R and Python
Show less

• Analyzed flight performance using data set from Bureau of Transportation Statistics
• Created a rich Dashboard and prepared User Stories using Tableau to analyze data of over a year
• Competitor Analysis – On-time performance for all the carriers with respect to American Airlines
• Recommendations - Provided recommendations to improve the efficiency of the flight operations
• Analyzed delay reasons in all the operating states of USA

Tools Used: Tableau

• Performed data analysis, reduction & transformation using Principal Component Analysis (PCA) technique for a dataset of 480 columns
• Computed lift and profit values for the data points using RapidMiner and predicted donors for the campaign to maximize the profits

Tools Used: RapidMiner and R

• Analyzed the Flixster network and computed the relevant network metrics
• Established the nature of the network(preferential or random) and then used community detection algorithms to detect a set of communities within the network
• Simulated an epidemic model on the generated communities to study the flow of information within the network
• Plotted the infection rate in these networks against that of randomly generated Preferential attachment network
• Predicted the pre-releasing… Show more

• Analyzed the Flixster network and computed the relevant network metrics
• Established the nature of the network(preferential or random) and then used community detection algorithms to detect a set of communities within the network
• Simulated an epidemic model on the generated communities to study the flow of information within the network
• Plotted the infection rate in these networks against that of randomly generated Preferential attachment network
• Predicted the pre-releasing movie time requirements for advertisement and provided insights for movie producer to gain more profits that saved around 25-30% of the advertising money

Tools Used: Gephi and R

Show less