Electrochemical impedance spectroscopy (EIS) is a common method in biosensing detection of pathogens for public health and safety. In its most general form, increases of charge transfer resistance or decrease of double layer capacitance at the interface are used for reporting EIS system changes due to pathogens. However, this strategy is not universally adaptable to various EIS sensors and could lead to inaccurate detection. Herein, we demonstrated a machine learning-based EIS biosensor for E.coli detection with improved accuracy. EIS data was obtained from gold electrodes immobilized with E.coli through antibody binding and fitted with the Randles model to extrapolate multiple impedimetric parameters. A machine learning model, using principle component analysis and support vector regression, was trained to automatically establish a quantitative relationship between multiple impedimetric parameters and bacterial concentrations. Results showed an improved accuracy in determining bacterial concentration. The improvement is due to the integration of both capacitance and resistance information. These results thus pave the way for automatic and accurate EIS biosensors in various applications.