In past several decades, gait biometrics has expanded rapidly and is now a viable alternative to traditional methods of identification. The technique used for gait surveillance, monitoring, and analysis is constantly evolving. In applications that use computer vision, it’s challenging to determine someone's gender by their stride. This study suggests a reliable and adaptable technique for determining gender based on gait to address this issue. More than ever, there is a requirement of dataset which is based on gait analysis and recognition that enables the extraction of useful data. Furthermore, most of research studies utilizing gait datasets and machine learning techniques are image-based. Thus, research on gait recognition has been conducted on the vast majority of available datasets. The OULP-Age dataset from OU-ISIR is used to evaluate the proposed system for gait analysis based on gender predication. It represents a person's gait using a gait energy image (GEI).Moreover, these GEIs images were used to extract the feature using pre-trained models and we used the XGBoost classifier to classify the gender prediction. After that, the XGBoost classifier’s parameters are tuned to improve the outcomes. Finally, DenseNet models achieved good results with use of XGBoost tune parameters.