This paper presents a framework for short-term travel time prediction in a motorway with a three-stage architecture: traffic flow forecasting, traffic flow generation and travel time extraction. Traffic flow forecasting reads the historical traffic data and utilizes a forecasting model - Autoregressive Integrated Moving Average (ARIMA) to predict short-term traffic flow. The traffic flow generation utilizes the Cell Transmission Model (CTM) to generate outgoing flow of a road of interest based on the predicted incoming flow from ARIMA. Predicted short term travel times can then be obtained through N-Curve Analysis. Compared to most studies, this paper presents a historical data-driven framework for travel time prediction that can be trained based on specific profiles of routes and cities. The motorway M4 in Sydney, Australia was used to test this framework. It is shown that the predicted travel times can be used to anticipate congestion episodes at the network level.