Proper intent classification of web queries is significant in providing users with accurate search results. For STEM-related searches, the generalized search engine provides some discrete results, and it becomes challenging to find the desired ones. Here, in our work, we have used a semi-supervised process and compared it with supervising approaches. This process has been done on Java and C# Bing web queries. From the performance comparison, we have found that our semi-supervised model has performed better than others according to accuracy and f1-score. We have also analyzed the performance by changing training data size, doing error analysis on all models, and finished by presenting how this prediction can be used on a search data fetching process.