Suparno Datta

Drive Personalised customer experience using data science.

Title - "Autoencoder networks for detection of fraudulent financial activities in ERP systems" Supervisor- Prof. Dr. Christian Bauckhage Goal- This thesis studies the suitability of different unsupervised anomaly detection techniques such as Repeated Neural Networks, Local Outlier Factor, K-Nearest Neighbors on Enterprise Resource Planning (ERP) data, primarily aimed at detecting fraud or potentially suspicious activities from a data mining perspective. Programming language used - R, Python

Activities- 1) Setting up and maintaing Storm clusters for streaming analytics. Used storm to perform twitter stream analysis ( sentiment analysis and text mining) 2) Predictive Analytics with R, Python & Vowpal Wabbit for retail marketing 3) Evaluating different BigData Technologies and tools in PwC BigData Lab ex.- Hadoop, Hive, Impala. 4) Applying and evaluation of different clustering algorithms on financial transaction data 5) Data visualization using Rshiny and Gephi

Activites- 1) Worked as a software developer. Mainly worked on optimizing already existing systems both in terms of time and space complexity. 2) Also briefly worked on implementing a few graph algorithms using LEDA.

Machine Learning, Data Mining, Pattern Recognition, Image Processing

Algorithms, Java, Distributed Systems, Networking