Mining products play a critical role in our current lives for instance titanium is used in the design of surgical pins and bone plates, renewable energy technologies, use copper for wiring of solar panels. Mobile phones are powered by precious metals for example lithium. To obtain this minerals, explosives are usually applied in rock fragmentation. Ineffective use of explosive energy in an operation may result in excessive ground vibration. Extreme ground vibrations because of blasting activities can result in various problems, for instance damage to property of the nearby residents and ecological damage. The measurement of ground vibration because of blasting is essential to mitigate risks associated with adverse impacts of blasting. The peak particle velocity (PPV) is the most important parameter generally used to evaluate ground vibrations in blasting sites. There are various methods that are currently used in prediction or estimation of PPV. The Empirical approach is limited by few input variables used in the prediction of ground vibration. The statistical and mathematical modelling techniques usually require explicit knowledge and understanding of the progression of the intricate blasting dynamics. With the current paradigm shift towards automated systems in the mining industry and introduction of the fourth industrial revolution concepts, it is imperative that other techniques for the prediction of PPV be applied. The main objectives of this research work is to conduct a study on the application various machine learning algorithms including deep Neural Networks in the prediction of PPV, 799 observations of data sets are used to develop machine learning algorithms. The following input features are used: the powder factor (kg/m3), spacing (m), stemming length (m), burden (m), maximum charge per delay (kg), blast-face distance to the monitoring point (m) and PPV is considered as the target variable. Random Forest, Artificial Neural Networks and deep Neural networks are some of the machine learning algorithms that are used in this research work. Various criteria, including mean absolute error (MAE), and correlation coefficient (R), are used to evaluate the developed models' accuracy and applicability. The results show the deep Neural Networks outperformed the traditionally known machine learning algorithm. To improve the performance further, the size of the data sets will be increased and various algorithms will be amalgamated.