Gears are important machine elements that transfer motion by the meshing of teeth. In this modern era gearbox has become an essential component in industry as well common man day to day life. In most of the industrial conditions gear boxes are subjected to continuous operation, which in many cases evades the maintenance activities. In such scenario the gear box may experience an unexpected failure, which will lead to shut down of the specific unit. Considering the significance of the gearbox, condition monitoring of gear box becomes essential. The helical gear box consists vital components like, helical gears of different ratios, bearing, shaft, gear shifting rod, plumber block etc. But the component which is prone to frequent failure must be prioritized and condition monitored, to ensure a continuous operation of the gearbox. That gives a scope for classification problem using machine learning algorithms. An experimental set was fabricated, and the vibration signals are acquired using a accelerometer sensor for the various faulty conditions like helical gear running in Good Condition (GC), Helical Gear with Tooth Crack Condition, (TCC), gear in Scuffed Condition (SC) combination of both in good condition (GC) and Scuffed Condition (SC). The signal acquisition system acquires the vibration signal under these conditions and fed into machine learning algorithms Naive Bayes and Support vector Machine (SVM). The Training Accuracy of signals of all three Sets under different Load conditions were acquired and accuracy of Naive Bayes was found to be around 92% which is far superior than any SVM Algorithm