An age-old challenge in space research is the search for an accurate prediction algorithm of auroral disturbances in the complex space environment system. Based on six months of observations from the Ultraviolet Imager on the Polar satellite, we have conducted a statistical study of four attributes of over 9000 auroral activity sites. These attributes are area, power, and their rates of change. The results provide us with a new technique to forecast auroral activity by setting up criteria on these activity attributes. The derived criteria for predicting 1 TJ total auroral energy dissipation at an activity site give a probability of correct high dissipation prediction of 0.85, a probability of correct low dissipation prediction of 0.95, a true skill score of 0.80, and an averaged advance warning time of 42 min. Similar prediction statistics are found for 0.1 and 10 TJ energy dissipation predictions. The prediction statistics indicate a high degree of forecast accuracy for this new technique.