Conventional methods have poor prediction accuracy and can not provide probability distribution characteristics of predicted results for great fluctuation and uncertainty of development indices in later period of oilfield development. A novel prediction method for development indices based on uncertainty research is put forward. This method firstly determines the probability distributions of the influencing factors of index through analyzing the historical production data.Then，a large number of samples for each factor can be produced by random algorithm based on the obtained probability distribution. Finally，the probability distribution of target index is predicted through establishing the quantitative relation between the index and its influencing factors. The injection-production ratio was predicted for H oil production plant by using this novel method. The prediction result for the first six months in 2013 shows that this method has higher prediction precision（the average error is 0.53%）compared to polynomial regression method（the average error is 3.33%）and support vector machine model（the average error is 1.46%）. The range of possible injection-production ratio in January 2013 is from 0.77 to 0.93，and the value of 0.834 3 is the most likely to occur. The novel method for development indices prediction based on uncertainty research provides more reliable basis for oil development decision-making，thus greatly lowers decision-making risk.