melikeoflu / 2018_brfss_survey_data_anlaysis

Developed and assessed 25 machine learning classification models using Python, pandas, scikit-learn, seaborn, and scipy.stats, effectively predicting various health conditions.

2018 BRFSS Survey Data Analysis

Utilizing Machine Learning Classifier Algorithms to Classify 2018 BRFSS Survey Data Prepared by the CDC

Required Libraries

• pandas
• scikit-learn
• numpy

Files Included:

• data folder: Contains source data and processed data.
• images folder: Charts and graphs used for project presentation.
• dataPrep: Documentation of the data cleaning process performed.
• data_processing: Details on attribute selection methods, model creation, and model evaluation.
• project_report: A summary report of the project.

Project Objective

The goal of this project is to build and evaluate classifier models using real-world data. This involves building 25 classifier models, utilizing various attribute selection methods. The models will be assessed based on a weighted average of TP rate, FP rate, precision, recall, and F-measure.

Project Outcome

• Successfully built and tested 25 classifier models using real-world data.
• Achieved an average accuracy score of 70% and a sensitivity rate above 50%.
• Employed data cleaning, preprocessing, feature engineering, and model selection techniques.
• Utilized Python, pandas, scikit-learn, seaborn, and other tools to optimize model performance and understand feature importance.

Project Steps

1. Data Cleaning:

   • Extensive data cleaning was required due to the real-world nature of the dataset.
   • Addressed issues such as populating missing values and replacing outliers with mean, median, or most frequent values.
   • Adjusted data types for analysis compatibility.

2. Data Processing and Model Creation:

   • Executed five iterations, each employing different attribute selection methods and classifier models.
   • Evaluated models based on accuracy, error rate, sensitivity/recall, and precision rate.

3. Best Model Selection:

   • Selected the best model based on average performance metrics and the efficacy of the feature selection method.