Volha Puzikava's repositories
Twitter_Sentiment_Analysis
Conducted Twitter sentiment analysis on Google and Apple products. Employed a range of supervised ML algorithms and neural networks to tackle an imbalanced NLP multiclass classification problem. Selected the model with the highest macro F1 score for evaluation.
Cryptocurrency_Prediction_Analysis
Forecasted price trends for the top two cryptocurrencies for a half year out starting September 2022. Detrended the data through subtracting the EWMA and differencing transformations. Modeled the data using various strategies including different orders of AR, MA, ARIMA, and SARIMA. Used the models with the lowest AIC for forecasting.
House_Sales_Price_Analysis
Developed an algorithm that predicted the best price for selling houses in the King County, Washington. Built various linear regression models based on statistically significant features, p-values and multicollinearity. Applied Recursive Feature Elimination and Brute Force approaches. Investigated linear regression assumptions.
Office_Supplies_Recommendation_System
Recommended office supplies based on reviews of purchased products and advised if it would be valuable to offer products as a two-pack. Worked in surprise library and Spark programming environment. Built recommendation systems using SVD and ALS models. Performed A/B testing to determine the probability of success for a two-pack.
Food_App_Analysis
A 1-Week Live Data Project for Data Scientist by HiCounselor. Included Preprocessing and cleaning the data using Python as well as running SQL queries to solve business problems.
ML_Algorithms_Course
Public repo for 365 Data Science ML Algorithms Course