This paper presents an xApp-based Machine Learning (ML) solution for dynamic resource allocation in Open Radio Access Network (O-RAN) systems. The O-RAN architecture enables disaggregated, distributed, and virtualized radio access networks, which support dynamic resource allocation. The proposed ML-based xApp optimizes resource allocation by dynamically adjusting the allocation of Physical Resource Blocks (PRBs) based on traffic demand and Quality of Service (QoS) requirements. The xApp uses a Random Forest Classifier to select the best allocation policy for each base station, achieving high accuracy (85%) in policy selection. The model is trained using O-RAN instance state input parameters and converges quickly to changing network conditions. The paper discusses the importance of dynamic resource allocation in 5G networks, where Network Slicing allows the creation of multiple logical networks to meet diverse service requirements. The Near-Real-Time RIC (near-RT RIC) plays a crucial role in monitoring and controlling network performance, while the xApp enables proactive resource allocation and auto resource allocation to ensure optimal utilization of available radio resources. The paper proposes four different resource allocation policies: Equal Allocation, Voice Priority, MBB Priority, and Dedicated Resources Reservation. These policies are evaluated through simulations to determine their impact on network performance, particularly in terms of user outage. The results show that the Dedicated Resources Reservation policy performs best under certain conditions, while the Equal Allocation policy is more effective in other scenarios. The ML-based xApp is trained using comprehensive simulations to generate datasets that include user counts, outage information, and network configurations. The Random Forest Classifier is chosen as the ML algorithm due to its high classification accuracy and efficient training time. The xApp enables dynamic selection of resource allocation policies based on network conditions and user demands, leading to improved network performance, user satisfaction, and Quality of Experience (QoE). The integration of ML within the xApp enables a responsive and efficient resource allocation process, allowing for the accurate estimation of current traffic demands and dynamic adaptation of PRB allocation. This approach ensures that network slices receive the appropriate allocation of resources, leading to improved user experience, optimized network efficiency, and enhanced fulfillment of Service Level Agreements (SLAs) and QoS requirements. The study highlights the transformative potential of ML in advancing network efficiency and performance, paving the way for more resilient and adaptable telecommunication networks.This paper presents an xApp-based Machine Learning (ML) solution for dynamic resource allocation in Open Radio Access Network (O-RAN) systems. The O-RAN architecture enables disaggregated, distributed, and virtualized radio access networks, which support dynamic resource allocation. The proposed ML-based xApp optimizes resource allocation by dynamically adjusting the allocation of Physical Resource Blocks (PRBs) based on traffic demand and Quality of Service (QoS) requirements. The xApp uses a Random Forest Classifier to select the best allocation policy for each base station, achieving high accuracy (85%) in policy selection. The model is trained using O-RAN instance state input parameters and converges quickly to changing network conditions. The paper discusses the importance of dynamic resource allocation in 5G networks, where Network Slicing allows the creation of multiple logical networks to meet diverse service requirements. The Near-Real-Time RIC (near-RT RIC) plays a crucial role in monitoring and controlling network performance, while the xApp enables proactive resource allocation and auto resource allocation to ensure optimal utilization of available radio resources. The paper proposes four different resource allocation policies: Equal Allocation, Voice Priority, MBB Priority, and Dedicated Resources Reservation. These policies are evaluated through simulations to determine their impact on network performance, particularly in terms of user outage. The results show that the Dedicated Resources Reservation policy performs best under certain conditions, while the Equal Allocation policy is more effective in other scenarios. The ML-based xApp is trained using comprehensive simulations to generate datasets that include user counts, outage information, and network configurations. The Random Forest Classifier is chosen as the ML algorithm due to its high classification accuracy and efficient training time. The xApp enables dynamic selection of resource allocation policies based on network conditions and user demands, leading to improved network performance, user satisfaction, and Quality of Experience (QoE). The integration of ML within the xApp enables a responsive and efficient resource allocation process, allowing for the accurate estimation of current traffic demands and dynamic adaptation of PRB allocation. This approach ensures that network slices receive the appropriate allocation of resources, leading to improved user experience, optimized network efficiency, and enhanced fulfillment of Service Level Agreements (SLAs) and QoS requirements. The study highlights the transformative potential of ML in advancing network efficiency and performance, paving the way for more resilient and adaptable telecommunication networks.