I have been working as a data analyst from past 3 years using Python, Tableau, SAS, Power BI, R Programming. I love solving business issues with historic data, Data Analytics is not just my profession it's my passion.
-
Experience
National Autobody Parts Warehouse Inc:
• Developed a machine learning model for competitive pricing analysis of autobody parts and to provide improved customer satisfaction by considering different certifications, vendors and insurers data which resulted in increase of sales by 12%.
• Creating ad-hoc KPI reports and interactive dashboards using Python and Tableau to analyze and optimize the company metrics on sales, purchases and logistic departments. This helped in reducing the total labor cost by 9% for logistics.
• Applying statistical techniques to identify clusters of products which can improve the cross-selling and upsetting profits.
• Managing the master data of all the departments and cleaned/processed data (ETL) using complex SQL queries.
-
Projects
Prediction and improvement of Contract Renewals:
Constructed a logit probability model in python to predict the contract renewals based on past service incident data using exponential time decay, correlation with multiplicative impact from the number of cases and escalations, determined maximum likelihood using optimize function in SciPy. Gathered the complete contract information by joining different datasets, pre-processed and cleaned the customer comments by analysing text using regular expressions, topic modelling with LDA to improve prediction accuracies. The final model was able to increase the contract renewals by 8%.
Sales Analysis with diminishing returns and Hypothesis Testing: Built a statistical model from scratch using diminishing returns and ran the model simulation data variables to check the robustness (89.5%) of the model. Found the effects of diminishing factor, dependency of interactions on the sales and determined the optimum coefficients using hypothesis testing and maximum likelihood as well as GridSearch algorithm with StratifiedKFold technique. The developed model is 18% better than the machine learning models on Scikit-Learn Library with respect to accuracy.
Chatbot:
Created a chatbot using September 2012 Reddit comments which works efficiently for introduction conversations. The JSON formatted massive data is stored in the MySQL database using the buffering technique and an ADAM optimized model is created using the seq2seq convertor from NMT (Neural Machine Translation) model in TensorFlow.