Malware detection and classification systems are essential components of cybersecurity. The identification and categorization of various unknown and variant malware strains present an ongoing challenge. Current literature indicates that visualization-based malware detection and classification models leveraging deep learning demand substantial datasets, high computational resources, significant storage space, and extensive training times. There is a pressing need to reduce these computational costs, storage requirements, and training durations while enhancing model performance to achieve greater efficiency. This paper contributes by proposing an improved malware classification model that employs 1D-CNN and stacked ensemble techniques. In our experiments, we utilized two sets of custom and classical 1D-CNN models as weak learners. The top-performing weak learners were selected to create two ensemble models. Among these, the best-performing ensemble model achieved a classification accuracy of 98.01% with minimal computational resources, significantly improving efficiency.