Microarray technology is widely used in cancer prediction and diagnosis. However, microarray data faces challenges such as small sample size, high dimensionality, and class imbalance, which pose difficulties in data analysis. This paper proposes an AdaBoost integration framework based on multiple filters (AIF) to alleviate classification complexity and improve classification accuracy. First, multiple filters are integrated to select features that represent diverse original information. Second, an iterative elimination strategy is proposed to address the class imbalance issue, ensuring that the learning process prioritizes features capable of accurately classifying minority class samples. Additionally, an iterative approach based on outlier detection is proposed to mitigate the influence of outliers on the learning process. Finally, the optimal subset is selected using a recursive elimination method. The proposed approach was validated on 5 microarray datasets. The evaluation metrics included three classification indicators (accuracy, F1, G-mean). Comparative experiments with other methods using two classifiers (SVM, KNN) demonstrated that AIF significantly improved classification accuracy and reduced the number of features selected.