Evaluation processes should focus on estimating the outcomes of social projects and interventions because good intentions and positive looking goals do not always lead to the achievement of assumed and intended social benefits. There are many finance-economics evaluation methods and models. Against this background, the achievements in the field of social aspects evaluation are much more modest. Expanding the set of available solutions, it is worth looking for sources of inspiration related to the digital era, new business analytics methods and machine learning approaches. The main purpose of this paper is to explore the possibilities of using Support Vector Machine (SVM) data mining for social project evaluation. The proposed method can be a significant complement to the solutions used so far in this field. Properly conducted evaluation of social projects is one of the most important conditions for their proper implementation and success. The method allows conducting research using large collections of various data, it can be a real support in reducing the subjectivity of evaluation processes regarding social aspects of interventions from public funds by obtaining reliable research results and relatively easy identification of calculation errors. The paper presents theoretical considerations as well as an example implementation of the system supporting the evaluation of innovative social projects and the results of empirical research related to the verification process of the proposed method. Possible directions of further research in the field of development of the presented solutions are also indicated.