Eureka! Institutional Repository

Missing values in datasets is a very important research issue in big data analysis. These datasets are often used for training machine learning models and if a significant percentage of the values is missing, it may result to inaccurate predictions or incorrect model evaluations. To address this issue, several imputation techniques have been proposed as part of the data cleaning process. However, applying these techniques to real-world datasets can be challenging and time-consuming for researchers and data scientists. This thesis presents the development of a web application that utilizes various imputation methods, offering an easy and user-friendly way to handle missing values in datasets. Users can access the website, upload their datasets with missing values in CSV format, choose one of the available imputation methods based on the feature types of the dataset and then download the file containing the imputed values, as soon as the imputation process is complete. The web application, named WEBIMPUTER, offers a variety of imputation solutions for numerical, categorical and mixed feature datasets, providing a wide range of parameter options for the imputation models. Finally, several experiments that have been conducted by applying all the imputation algorithms of the application to various datasets of different file size and measuring the execution time are presented here, to help users gain a better understanding of the computational efficiency of the models.