Service component selection is a core problem in software development process. With an enormous number of components available, it is often difficult for the developer to select the most appropriate one, as he or she might not be aware of all the possible business scenes ahead of time. Taking these challenges into consideration, we propose a deep learning-based system that automatically recommends service components based on component selection history during the software development process. We employ a sequential model with two long short-term memory (LSTM) layers and two fully connected layers, using SoftMax as an activation function, to predict the next service component. The model was trained, validated and tested on dataset with more than 120,000 examples from a real-world software company. The proposed network outperforms the baseline methods in terms of the evaluation criteria. In addition, the model results were deployed in a real-world software tool and gave positive feedback.