As a data scientist at a retail company, you are tasked with training a sales prediction model using a managed tabular dataset in Vertex AI. The dataset contains sales data from three different stores, including features such as store name and sale timestamp. The goal is to make accurate sales predictions for a new store that will open soon. To achieve this, you need to split the data between training, validation, and test sets. Which approach should you take to split the dataset effectively?