Geo Data Science

Prem George Alphonse

Masters of Data Science

RMIT University, Melbourne

• R, Python, SAS, JAVA, SQL, Machine Learning, Data Visualisation, Data Modelling, Gephi, Carto, Excel – Pivot Tables, Hadoop Big Data Processing using AWS Elastic MapReduce.
• Data Science, Machine Learning, Time Series Analysis, Social Media and Network Analytics, Data Visualisation, Statistics, Mathematics, Programming.
• Data Science Process, Data Modelling on Classification, Clustering and Prediction Analysis.
• Supervised Learning Techniques: KNN, Decision Tree, Regression, Random Forest, Naïve Bayes, SVM, Neural Networks.
• Unsupervised Learning Techniques: k means, DBSCAN.
• Recommender System, Collaborative Filtering.
• Social Media and Networking Analysis: Graphs, Community Detection, Sentiment Analysis, Information Cascade, Topic Modelling, Recommendation.
• Query Suggestion: Agglomerative clustering of click data, Automatic relevance feedback.
• Machine Learning Libraries: Data Wrangling – Pandas, Visualisation - Matplotlib, Seaborn, Machine Learning - Scikit-Learn.
• R Language: Descriptive Statistics Visualisation, Estimating Uncertainty, Hypothesis Testing, Simple Linear, Logistic, Poisson, Negative Binomial,Quasi-Poisson Regression and Correlation, Time Series Modelling (ARIMA, SARIMA, ARCH, GARCH), Machine Learning Algorithms.
• Data Visualisation: Craft and Grammar, Visual Perception, Univariate, Bivariate and Multivariate Data Visualisations, Mapping of ggplot and Spatial Data Visualisation (Carto), Interactive Visualisation (Plotly), Apps for Data Visualisation (Shiny).
• Former member of Engineers Australia (Overseas Engineering Qualification accessed by EA)

LinkedIn Recommendations

Master of Data Science Test Scores Analysis Using Machine Learning

Instead of just saying "Completed Masters of Data Science", let's test the study by its own medicine and evaluate the performance. Program gave me the Class labels as grades on my courses like High Distinction, Distinction..., unfortunately i got only these two. Created Classification Models and justified the grades. I have a micro mini dataset of my courses, 15 records, class is imbalanced, its "unfortunate" that i got more HD's (11 HD and 4 D), so used Synthetic Minority Oversampling Technique to balance it, did feature selection using information gain, gain ratio and chi-squared, and selected features using chi-squared which gave more importance, and created fused learners with AdaBoostM1, LogitBoost, RandomForest, Multilayer Perceptron and Naive Bayes. I am more interested in my False Negatives and i wish zero False Positives, all my models gave me Zero FP, i wished 2 FN's, in two more of my courses i felt deserved to get HD's, Naive Bayes didn't help me, Neural Networks said me, you can have One, but i liked AdaBoostM1, LogitBoost and RF, they said here you go with 2, based on voting with 3 classifiers, i accepted it. and i liked Matthews Correlation Coefficient which considers all and gave me 0.65 in all three classifiers.

Certifications