XGBoost is probably the best model for tabular data with independent observations (not time-series). It's based on an ensemble of decision trees and won a lot of public competitions. XGBoost supports classification, regression, count, and survival modes.
In many cases, a dataset is too small to evaluate a model performance correctly. A standard practice of splitting data on training and test sets gives unreliable results. Cross-validation helps with this problem by running multiple splits iteratively and calculating a more robust score based on those runs.
After a model is fit and made predictions, you can download a resulting CSV file for further usage. It includes the same input columns together with the predicted target column. You can also download the fitted model in JSON format and charts with XGBoost performance on this dataset as SVGs.
XGBoost is a native C++ library ported to WebAssembly and wrapped in this web application. When you load a dataset, it's processed using a virtual machine running in your browser. That means no single byte of your data is leaving a device that runs this web app.