Human-AI Pair Programming System: A Case Study on Railway Congestion Forecasting

We are developing a system called SynergyAI based on a novel concept of human-AI pair programming. This concept does not focus solely on advising human programmers by AI but also stimulates the programmer's creativity and AI optimization ability. In human-AI pair programming, the human first chooses the data and AI operators. Then, the AI optimizes the parameters based on the correlation of the data. The output from the data is the prediction result and prediction factors. As a case study, we programmed a system to forecast railway congestion. In the case study, there are many parameter stations and spots along the railway line. The human programmer decides these parameters, producing an optimized program with suitable parameters. Based on human congestion factor input, the AI optimizes the relation of data to achieve high forecasting accuracy. As a result, we can predict the congestion factor with more than 90 % accuracy and about 70% improvement in terms of the complexity of the data flow. This shows the effectiveness of the pair programming system.