Conclusive Report

Data Wrangling

1. Initial Data Overview:

   • The dataset contains 520 rows and 17 columns.
   • Columns include demographic and symptom-related information, with a target class indicating diabetes status.

2. Missing Values:

   • No missing values were found in the dataset.

3. Duplicate Rows:

   • 269 duplicate rows were identified and removed, leaving 251 unique rows.

4. Data Types:

   • 'Age' is an integer, and all other columns are categorical (object type), which is appropriate for the data.

Data Analysis

1. Distribution of Age:

   • The age distribution was visualized using a histogram.
   • The dataset contains a wide range of ages, with a noticeable concentration in the middle age group.

2. Gender Distribution:

   • The gender distribution was visualized using a count plot.
   • The dataset has a balanced distribution of male and female participants.

3. Diabetes Class Distribution:

   • The distribution of diabetes class was visualized using a count plot.
   • The dataset contains a higher number of positive diabetes cases compared to negative cases.

4. Distribution of Key Features by Class:

   • The distribution of key features such as 'Age', 'Polyuria', 'Polydipsia', 'sudden weight loss', and 'weakness' was visualized by diabetes class.
   • These visualizations help in understanding the relationship between these features and the target variable.

Feature Importance

1. Random Forest Feature Importance:

   • The feature importance was calculated using a Random Forest classifier.
   • The top 5 most important features are 'Polyuria_Yes', 'Polydipsia_Yes', 'Age', 'Gender_Male', and 'partial paresis_Yes'.

2. Logistic Regression Coefficients:

   • The coefficients of the logistic regression model were analyzed.
   • The features with the highest positive coefficients are 'Polydipsia_Yes', 'Polyuria_Yes', and 'Genital thrush_Yes'.
   • The features with the highest negative coefficients are 'Gender_Male', 'Itching_Yes', and 'delayed healing_Yes'.

Model Performance

1. Model Comparison:

   • Several models were trained and evaluated, including Logistic Regression, Random Forest, Decision Tree, SVM, and KNN.
   • The Random Forest model achieved the highest accuracy of 92.16%.
   • The SVM model had the lowest accuracy of 68.63%.

2. Logistic Regression Performance:

   • The logistic regression model achieved an accuracy of 72.55%.
   • The precision, recall, and F1 score were also calculated to evaluate the model's performance.

Summary Statistics

1. Age Statistics by Class:
   • The summary statistics for age by diabetes class were calculated.
   • The mean age for negative cases is 46.36 years, while the mean age for positive cases is 49.07 years.

Conclusion

• The dataset was successfully cleaned and analyzed, with no missing values and duplicate rows removed.
• Key features influencing diabetes status were identified using feature importance and logistic regression coefficients.
• The Random Forest model was the best-performing model, achieving the highest accuracy.
• Further analysis can be conducted on the cleaned dataset to explore more complex relationships and improve model performance.