Currently，many studies dedicated to context aware based recommendation，considered different types of context properties，but they ignore the important degree of different context attribute impact the behavior，that is，context strength．This paper defines the customer context，context intensity，and behavior changing quantitatively; presents context strength constrained pattern mining methods and change detecting method to extract the critical situation caused by changes in behavior．The proposed algorithm increased the sensitivity of the interests changing，improvement of the massive data under support for sparse association rules，the shortcomings of low confidence sensitivity．Experiments and analysis demonstrated the feasibility and effectiveness．