In order to realize Digital Oil Field, some key problems need to be improved, esp. accurate and automatic prediction of oilfield development indexes which may be resolved by designing of intelligent prediction system. With the shortcoming of inference of system designed by us, automatic inference problem for a complicated intelligent prediction system was improved using pattern recognition method. First, intelligent prediction system and the methods as well as principles of pattern recognition were introduced. Then the framework of intelligent prediction system based on pattern recognition was formulated by using technologies and methods of human-computer interface, fuzzy processing and pattern recognition. Secondly, the knowledge base was extended as augmented knowledge base with introducing credibility to measure uncertainty of knowledge. Particularly, the methods and principles of pattern recognition were used to design two recognizers and one inferring machine. Moreover, the method of selecting predictive model based on reasoning of pattern recognition was presented by coupling them and intelligent prediction system. Finally, the design of improving intelligent prediction system of oilfield development indexes was simulated. Simulation result shows that improved system may automatically realize to select optimal prediction model by computer according to different reservoirs and different development stages. The results obtained in this thesis will helpful to design for intelligent prediction system,