State of charge estimation for lithium-ion batteries in electric vehicles is the primary aspect of this research. The traditional models for state of charge estimation have been compared alongside the modern machine learning methods. An actual BMW i3's testing data has been considered for the practical validation of the research and to further validate it, a simulation-based dataset from MATLAB Simulink has also been tested parallelly. The actual practical physical methods like Coulomb counting and open circuit voltage have been considered. And four machine learning models, namely random forest, support vector machine, K-nearest neighbor and linear regression have been used. The results obtained display a better result for the machine learning models particularly random forest over physical models for state of charge estimation. A very low mean absolute error was found and a nearly perfect Pearson coefficient which shows the model's quality of performing well even on nonlinear dynamics of battery.