Android is the most widely used operating system today and occupies more than 70% share of the smartphone market. It is also a popular target for attackers looking to exploit mobile operating systems for personal gains. More and more malware are targeting android operating system like Android Banking Trojans (ABTs) which are widely being discovered. To detect such malware, we propose a prediction model for ABTs that is based on hybrid analysis. The feature sets used with the machine learning algorithms are permissions, API calls, hidden application icon and device administrator. Feature selection methods based on frequency and gain ratio are used to minimize the number of features as well as to eliminate the low-impact features. The proposed system is able to achieve significant performance with selected machine learning algorithms and achieves accuracy up to 98% using random forest classifier.