Tobacco Use Survey Data Analysis

Dataset Description

This report delves into the intricate world of tobacco use, examining trends across various demographics and geographical regions. By analyzing a comprehensive dataset, we uncover valuable insights that can inform public health strategies and interventions.

• YEAR: The year the data was collected.
• LocationAbbr & LocationDesc: Abbreviated and full state names.
• TopicType & TopicDesc: General and specific topic categories (e.g., “Tobacco Use – Survey Data” and “Cigarette Use (Youth)”).
• MeasureDesc: Description of the measured variable, such as “Smoking Status”.
• Response: The type of response, such as "Ever," "Current," etc.
• Data_Value_Unit & Data_Value_Type: The unit and type of the data value (e.g., "Percentage").
• Gender, Race, Age, and Education: Demographic details.
• DisplayOrder: Numeric order for display purposes.

Summary of Findings

• Shifting Patterns: We observed significant fluctuations in smoking rates among different age groups and educational levels over the years.
• Geographic Disparities: Distinct regional variations in tobacco use were identified, highlighting the need for targeted interventions.
• Demographic Influences: The impact of demographic factors such as gender, race, and age on tobacco use was explored, revealing nuanced patterns.

Data Preprocessing

1. Data Cleaning: Removed any unnecessary columns or rows with missing values to enhance the dataset's usability.
2. Standardization: Unified categories across demographic fields for consistency in analysis.
3. Transformation: Transformed certain columns (e.g., converting percentages to numeric values for analysis).
4. Encoding: Encoded categorical variables for model compatibility.

Exploratory Data Analysis

To visualize the trends and patterns within the data, we employed a variety of visualization techniques.

Visualization

1. Trends in Tobacco-Related Measures Over Time (Line Chart):

   • This line chart illustrates changes in tobacco-related measures (e.g., "Smoking Status," "User Status") across the years. The lines for each measure reveal patterns or shifts over time, indicating trends in smoking behaviors or tobacco use within the population.

2. Distribution of Tobacco Use by Age and Measure Description (Count Plot):

   • This count plot visually represents the distribution of tobacco use behaviors across different age groups, categorized by various types of measurement descriptions. By illustrating how many individuals report specific tobacco use behaviors at different ages, the plot provides valuable insights into trends in tobacco consumption.

3. Heatmap of Data Values by Education and Topic (Heatmap):

   • This heatmap shows the relationship between educational attainment and various tobacco-related measures across topics. Each cell's color intensity indicates the average value of the tobacco-related measures for each educational level, helping visualize how education may impact engagement with tobacco issues.

4. Combined Analysis of Tobacco Use Trends by Gender Over Time (Facet Grid):

   • This facet grid shows tobacco use trends over time, with separate histograms for each gender. Each subplot represents either male or female, and the stacked bars display the annual distribution for each tobacco measure, revealing any notable gender differences in smoking behaviors over the years.

Through this data-driven exploration, we have gained valuable insights into the factors influencing tobacco use. By understanding these patterns, public health officials can develop targeted interventions to reduce tobacco consumption and promote healthier lifestyles.

Model Development

To gain predictive insights, we developed several machine learning models:

• Logistic Regression: Used to model binary tobacco use outcomes.
• Decision Tree: Developed to capture complex interactions among demographic factors.
• Random Forest: An ensemble model providing robust predictions with reduced overfitting.
• Gradient Boosting: A sequential ensemble model that builds models in a stage-wise fashion, correcting errors from prior models to increase overall accuracy.

Model Evaluation

Each model was evaluated using metrics such as accuracy, precision, recall, and F1-score:

1. Logistic Regression:
   • Accuracy: 1.0 or 100%
   • Evaluation: Logistic Regression achieved perfect classification, with an accuracy of 100%, suggesting that the model was able to correctly classify all instances. This indicates clear class separation in the dataset, which this model handled effectively.
2. Support Vector Machine (SVM):
   • Accuracy: 1.0 or 100%
   • Evaluation: SVM achieved 100% accuracy, similar to Logistic Regression. The SVM model’s perfect precision and recall across all classes demonstrate its effectiveness at handling high-dimensional data with clear class separability.
3. Random Forest:
   • Accuracy: 1.0 or 100%
   • Evaluation: The Random Forest model also achieved perfect accuracy, precision, and recall. Its ensemble nature likely contributed to capturing complex relationships, with no overfitting observed in the test data. The model’s ability to generalize well is evident from the consistent metrics.
4. Gradient Boosting:
   • Accuracy: 1.0 or 100%
   • Evaluation: Gradient Boosting achieved a perfect score, similar to other models. By sequentially learning from prior errors, it effectively classified all test data points, demonstrating high adaptability and precision in handling challenging classification tasks.

We rigorously evaluated the performance of these models using metrics like accuracy, precision, recall, and F1-score. These metrics provide a comprehensive assessment of the models' ability to accurately predict tobacco use.

Conclusion

This analysis demonstrates significant insights into tobacco usage patterns across demographic groups and locations. Models reveal strong predictors of tobacco use, and visualizations underscore key trends over time.

Contributors

• Mike Fernando Bunag
• Mitch Dela Cruz