Ammonium nitrate based explosives are a choice weapon for many terrorist groups due to its ease in manufacturing and high velocity of detonation. These explosives undergo thermal decomposition to release ammonia gas in traces of about 5- 25 Parts per Million (PPM) below the olfactory threshold. Ammonia is a reducing gas. MQ137 sensors are low cost commercially available metal oxide semiconductor ammonia gas sensors with a problem of selectivity (reacting with other reducing gasses like carbon monoxide etc) and sensitivity. We present the optimization of MQ137 metal oxide semiconductor electrochemical sensor using MATLAB, to improve its selectivity and sensitivity for accurately recognizing the characteristics of ammonia gas within specified PPM range as a sign of ammonium nitrate based explosives in vehicles. In this study, MQ137 sensor was connected with an ARDUINO microcontroller to a digital computer (2.40 GHz processor) and pre-heated for 12 hours before being exposed to ammonia gas in a controlled environment at room temperature to extract features (sensitivity constant and concentration in PPM) of ammonia gas with MQ137 sensor. 150 data samples of each feature were extracted and trained in a multilayer pattern recognition neural network with one hidden layer and 50 data samples containing features of other reducing gasses from the data sheet were used for testing. Test performance of multilayer artificial neural network has an accuracy of 100% with no misclassifications.