One of the significant factors affecting customer satisfaction with Internet Protocol Television (IPTV) systems is zapping delay when changing to another channel. This paper describes a new framework, called the Predictive Burst-Assisted Channel Switching (PBACS) framework, intended to reduce the delay associated with changing channels. The PBACS framework combines the predictive ability of machine learning to predict user viewing behavior and multicast burst streaming capability to pre-load probable next-viewed channels prior to changing to the new channel. Using simulation tools such as MATLAB/Simulink, results showed that the PBACS framework reduces zapping delay time from a base of 150-210 milliseconds to 90-130 milliseconds; an overall average improvement of 36.7%, exceeding improvements made by the hybrid method by 11.4%. PBACS improved the best-case zapping delay from 150 to 90 milliseconds (40%) and the worst-case zapping delay from 210 to 130 milliseconds (38.1%). Additionally, resource costs declined by 34.2% and prediction accuracy improved to 85-95%, an improvement of 13.3% when compared to the baseline. Moreover, according to results, improvements in bandwidth efficiency (14.1%), Quality of Experience (QoE) (26.7%), system utility (27.5%), and user satisfaction (21.8%) also were achieved. Overall, findings indicate PBACS provides an efficient, scalable solution for providing live and on-demand IPTV services. Findings also indicate significant improvements in responsiveness of IPTV and improved user experience.