Establishing correlations between various properties of alloys and their compositions and manufacturing process parameters is of significant interest to materials engineers. Both physics-based as well as data-driven approaches have been used in pursuit of this. Of various properties of interest, fatigue strength, being an extreme value property, had only a limited amount of success with physics based models. In this paper, we explore a systematic data driven approach, supplemented by physics based understanding, employing various regression methods with dimensionality reduction and machine learning methods applied to the fatigue properties of steels available from the National Institute of Material Science public domain database to arrive at correlations for fatigue strength of steels and present an assessment of the residual errors in each method for comparison. This study is expected to provide insights into the methods studied to make objective selection of appropriate method.