Attribute weights exert a significant effect on the solution in multi-attribute decision analysis (MADA), since solutions produced by varying attribute weights probably vary. When a decision maker has inadequate valid data, understanding or experience to produce exact attribute weights, he/she perhaps wants to seek a solution with highest reliability, referred to in this study as a highly reliable solution. To this end, a high-reliability evidential reasoning (ER) approach is put forward in the present work, which achieves alternatives comparison through determination of their reliability relative to attribute weights under ER scenario. Initially, the best alternative supported by single or multiple sets of attribute weights was determined. Then, reliability estimation is given for every alternative. In the case of highest reliability, the optimal interval of attribute weights and evaluation grades between the optimal alternative is measured and their ranking is generated. The proposed approach to the process is based on a combination of identifying these alternatives and measuring their reliability. The problem of automobile performance evaluation is explored, finding that the proposed approach is capable of effectively generating high reliability solutions for MADA problems.