Published Paper Details:

Ensuring the quality of tabular data is a critical preprocessing step before training reliable predictive models. We present a web-based application built with Django that automates data quality assessment and cleaning on uploaded CSV datasets, followed by training of a supervised learning model. The system ingests user-provided CSV files and analyzes issues such as missing values, outliers, and skewed features, providing visual dashboards summarizing key data quality metrics. Automated cleaning operations (mean/mode imputation, outlier clipping via the IQR rule, log-transform for skewed features, and one-hot encoding of categoricals) are performed using Python libraries (Pandas, SciPy, scikit-learn), producing a cleaned dataset ready for analysis. Users can then select a target variable and train multiple regression/classification algorithms (e.g. linear regression, random forest, k-NN), with the best model highlighted by evaluation metrics (e.g. mean squared error). In a case study on a housing price dataset, the application identified approximately 5.97% missing values, 1.47% outliers, and 25.0% highly skewed features, yielding a composite data quality score of 67.56%. A Random Forest regressor achieved the lowest error among models, consistent with known performance. The interface displays feature correlations and per-unit effect sizes, along with interactive charts. This integrated solution streamlines the machine learning pipeline by combining statistical data cleaning and model training in a single Django-based framework.