Roshan Sureen

– Challenges and solution to Machine Learning-Based Emotion Analysis of Textual Data – Research involves creating a comprehensive comparison framework – Implementation of different algorithms for analysis and feature extraction – Comparing the algorithms performance with best set of features based on certain metrics – Providing drawbacks and challenges of each method – Developing a prototype for users to choose different algorithms and different features, to get results in a comparable way.

– Research Project as part of course Advanced Practical Data Science – Used NEW-CLS dataset. Defect classification of 6 classes of defects – Performed a comparative study of three DL algorithms; A vanilla CNN, ResNet50 and VGGNet16 – Transfer learning methods outperformed traditional CNN with accuracy ranging between 95% - 99% – ResNet50 provided better balanced performance

– Learned to apply software engineering methods & tools – Gained practical experience in building software project – Developed an AR remote controlled car game. – Tested the app during 'Night of Science' in Göttingen – Users were provided with a static on-screen menu version of the game and the other with AR menu – Data analysis showed that users did enjoy the change in menu perspective – Research yielded that the menu type as such does not immerse the user completely in the AR experience

– Data Analyst/ consultant for startup Happy Fridge – Data analysis performed on their 1 year old dataset – Data processing using pandas & Data Visualization using matplotlib – Optimised Product assortment & better decision making based on model performance

– Main concept - model a decentralized platform for the car insurance industry – Developed and deployed a Decentralized application on a private Ethereum network – Connected the app to IPFS via infura for storage of nonchain information – This lead to an optimized and non-bloated blockchain – Achieved a determination for the thesis

Data Science, Grundlagen der Wirtschaftsinformatik