Modern automotive engine ignition system includes many parts such as spark plugs, ignition wires, distributor cap, distributor rotor, distributor, ignition coil, etc. Many malfunctioning problems caused by the above components will affect the engine performance or even make the engine cannot start up. Currently, automotive mechanic uses an oscilloscope /scopemeter for diagnosing the engine spark ignition problem because there are well-defined signal patterns for diagnosing the spark ignition problems, The mechanic uses this method to capture the signal patterns and find out what may be the potential malfunctioning parts of an automotive engine. However, this diagnosing method may not be very precise because many spark ignition patterns are very similar and sometimes even an expert mechanic may be wrong. Therefore, many trials for diagnosis are necessary and every trial requires the mechanic to disassemble and assemble the engine parts, which makes traditional manual diagnosis very time consuming. A machine learning technique of Support Vector Machines (SVM) is proposed to improve the current spark ignition diagnostic method because it is well known that SVM can serve as a powerful tool for resolving the pattern recognition and classification problem. This technique can be used to classify those spark ignition patterns more accurately, Therefore, applying SVM on the spark ignition problem can significantly reduce the number of trials a mechanic needs to take and make the diagnosis more precisely and quickly In this research, SVM is employed to construct a model that facilitates the spark ignition pattern classification, The techniques using in our experiments and the details model setup are discussed, In addition, a multilayer feedforward networks (MFN)technique is also employed to build another model for comparison with the SVM model. The comparative analysis shows that the SVM model we purposed provides a better accuracy than MFN.