Friction stir blind riveting (FSBR) is a recently developed manufacturing process for joining dissimilar lightweight materials. The objective of this study is to gain a better understanding of FSBR in joining carbon fiber-reinforced polymer composite and aluminum alloy sheets by developing a sensor fusion and process monitoring method. The proposed method establishes the relationship between the FSBR process and the quality of the joints by integrating feature extraction, feature selection, and classifier fusion. This study investigates the effectiveness of lower rank tensor decomposition methods in extracting features from multi-sensor, high-dimensional, heterogeneous profile data. The extracted features are combined with process parameters, material stack-up sequence, and engineering-driven features such as the peak force to provide rich information about the FSBR process. Sparse group lasso regression is adopted to select the optimal monitoring features. The selected features are fed into weighted classification fusion to estimate the quality of the joints. The fusion method integrates five individual classifiers with optimal weights. The correct classification rates resulted from various feature extraction and selection methods are assessed and compared. The proposed method can also be applied to other manufacturing processes with online sensing capabilities for the purpose of process monitoring and quality prediction.