Artificial Intelligence-based green computing has emerged as a promising approach to maximize energy efficiency while leveraging the power of artificial intelligence. AI-based techniques enable the development of intelligent controllers and schedulers that optimize the execution of tasks and reduce idle time. Furthermore, AI-based green computing can benefit from advanced data analysis techniques, such as dimensional reduction (DR) to improve energy efficiency. DR techniques aim to reduce the dimensionality of the input data while preserving its important features. By reducing the amount of data processed, DR methods can significantly reduce energy consumption during training and inference stages. This study aims to bridge this gap by proposing and evaluating dimensional reduction in the context of image-based classification tasks, respectively. AI-based green computing offers several advantages over traditional approaches. One major advantage is the ability to leverage machine learning algorithms for intelligent decision making. In this study, we aim to bridge the existing gap in energy-efficient approaches to AI-based green computing. By delving into the challenges and limitations of Base Models, we will explore the possibilities offered by dimensional reduction technique. In the critical field of medical diagnosis, particularly the classification of brain tumors the DR model with feature learning maintains nearly the same level of accuracy and recall as the base model but with greater efficiency.