This work presents a novel hybrid machine learning method designed to improve breast cancer diagnosis and prediction. The proposed method leverages the Wisconsin Diagnostic Breast Cancer (WDBC) dataset, which consists of 569 samples with 32 characteristics. It incorporates multiple machine learning models, such as Random Forest, AdaBoost, and Long Short-Term Memory (LSTM). In addition to feature selection approaches like Information Gain Ratio (IGR) and Chi-square tests, data pretreatment techniques like handling missing values and managing outliers are applied to guarantee robust model performance. To increase diagnosis accuracy, the hybrid approach cleverly blends the LSTM's analytical capabilities with the complimentary advantages of classical models. Confusion matrices and ROC curves are examples of graphical representations that facilitate the use of performance evaluation measures like precision, accuracy, recall, and F1 score. Outcomes indicate promising accuracy rates in breast cancer prediction, underscoring the potential for early diagnosis and improved clinical outcomes. This research contributes to advancing cancer diagnosis through machine learning techniques, promising advancements in personalized medicine and healthcare.