CS Megha Bisht

Certified Data Science professional with hands-on experience in projects involving machine learning, data visualization, and statistical analysis. Skilled in Python, Pandas, Scikit-learn, and SQL. Passionate about extracting insights from data and solving real-world problems. Eager to contribute to data-driven decision-making in a collaborative team environment.

💳 Credit Card Consumption:

Business problem-
A leading bank wanted to predict future credit card spending by customers to optimize credit limits, reduce risk, and enhance marketing strategies.

Objective-
To build a machine learning model that forecasts future credit card consumption using historical transactions and demographic data.

Data availability-

· Customer transaction history

· Demographic details like age, income, location

Tools used-

· Python: Pandas, NumPy for data processing, Matplotlib and Seaborn for visualization

· Machine Learning: Linear Regression, Random Forest, Gradient Boosting

Metrics-

·R² score (achieved over 80%)

Tuning parameters-
Used GridSearchCV for model tuning:

· Gradient Boosting: learning_rate, n_estimators

Challenges-

· Handling missing values and imbalanced features

· Avoiding multicollinearity

💬 Customer Review Analysis :

Business problem-
Understand customer sentiment and feedback patterns across products, categories, and demographics to improve product offerings and enhance customer satisfaction.

Objective-
Analyze and classify customer reviews to identify sentiment polarity and key drivers behind product recommendations.

Data availability-

· Customer reviews (text data)

· Associated metadata: category, sub-category, product name, customer age, and location

Tools used-

·     Python libraries: NLTK, Scikit-learn for NLP and modeling, WordCloud, Matplotlib

·     Text preprocessing: Tokenization, Stopword removal, Lemmatization

Metrics-

· Sentiment classification accuracy: Achieved over 77%

Challenges-

· Handling noisy and unstructured review text

· Balancing uneven sentiment classes (e.g., more positive than negative reviews)

⚡ Electricity Demand Estimation:

Business problem-
Forecast electricity demand to support energy providers in making data-driven supply, pricing, and capacity planning decisions.

Objective-
Estimate future electricity consumption using historical demand data and time-related features to improve forecast accuracy and resource planning.

Data availability-

· Historical electricity usage data

· Time-based variables: hour, day, month, season, holiday indicators

Tools used-

· Python: Pandas, NumPy for preprocessing, Matplotlib, Seaborn

· Time series modeling: ARIMA, Facebook Prophet

Metrics-

· MAPE (Mean Absolute Percentage Error) for forecasting performance

Tuning parameters-

· ARIMA: p, d, q values via ACF/PACF plots and auto_arima

· Prophet: changepoint prior scale, seasonality mode

Challenges-

· Ensuring stationarity of time series data

· Capturing multiple seasonalities (daily, weekly, yearly)

Certified Data Analyst – AnalytixLabs