In geochemical exploration, how to choose indicator elements and how to synthesize the concentrations of indicator elements into metallogenic favorability are two key problems that need to be solved properly in mineral exploration targeting. Developing new geochemical data processing techniques is one of feasible ways to solve these two key problems. Therefore, a recursive indicator elimination (RIE) method was proposed in this paper to simultaneously determine the optimal subset of geochemical elements and to complete mineral exploration targeting. The method is a combination of a recursive elimination process and machine learning techniques. The recursive elimination process is an iterative procedure to determine the optimal subset of geochemical elements; and the machine learning techniques are used to model the relationship between the subset of geochemical elements and the mineral deposits in each iteration. The optimal subset of geochemical elements determined by the recursive indicator elimination has the closest relationship with the mineral deposits. Thus, it is expected that the machine learning model established on the optimal subset of geochemical elements will have the best performance in mineral exploration targeting. A case study was conducted in the Helong area, Jilin Province, China. Either the recursive indicator elimination or the area under the curve (AUC) was used to determine the optimal subset of geochemical elements based on the 1:50, 000 stream sediment survey data. In addition, principal component analysis (PCA) was used to synthesize all geochemical elements into a few principal components (i.e., comprehensive indicators). Finally, the optimal subset of geochemical elements and the principal components were used to establish the logistic regression model for mineral exploration targeting. The results show that the RIE combined with logistic regression has the best performance in mineral exploration targeting. The performance of AUC combined with logistic regression is similar to that of PCA combined with logistic regression. From the geological point of view, there is a strong consistency between the mineral exploration targeting results and the metallogenic characteristics of the study area. Therefore, the combination of RIE and logistic regression is an effective method to synthesize geochemical element concentrations into metallogenic favorability for mineral exploration targeting.