Software is transforming the world by its speed, availability, reliability, and efficiency. Hence development of software needs a lot of precaution to avoid measure harm. This paper has developed a model that finds the defects in the software as per the number of lines of code, classes, functions, etc. This defect detection improves reliability and reduces the development cost. Many of the features were collected to find the defect out of effective feature selection was done by using of elephant herd optimization algorithm. Selected features were normalized and used for the training of a spiking neural network. Results show that the proposed Software Defect Detection by Elephant Herd Features Learning has enhanced the comparison of parameter values.