R&D alliance plays an important role in this fast changing business environment. Companies usually establish several R&D alliances to address the challenges brought by the short technology cycle time and heavy competition. A central and critical factor of the success of R&D alliance is the appropriate selection of partners. However, existing studies incur several limitations. In response, we adopt the patent portfolio mining approach to develop a partner selection for R&D alliance technique, which encompasses technological and innovative variables as prediction variables. Thirty technological and innovative variables derived from patent portfolio mining are included and the Naïve Bayes algorithm is employed as our underlying learning algorithm. Moreover, we collect real R&D alliance cases in the biotechnology industry in U.S. in the recent five years to evaluate our proposed partner selection for R&D alliance technique. The evaluation results are encouraging and will serve as a basis for future studies.