Smoke detection is essential for safety and fire protection systems, and incorporating machine learning (ML) algorithms significantly improves its precision and effectiveness. The ML techniques for binary classification are investigated and assessed in this work by utilizing different algorithms such as: Logistic Regression (LR), Naïve Bayes (NB), K-Nearest Neighbors (KNN), Decision Tree (DT), Random Forest (RT), and Support Vector Machine (SVM). The smoke detection dataset chosen for this study contains around 62,630 with 14 features instances where 44,757 instances have been identified as fire, whereas 17,873 instances have been classed as no fire. Moreover, these cases are determined to be unbalanced. The data pre-processing techniques utilized for training and performance evaluation are SMOTE-Tomek, the removal of unnecessary features, and the correlation matrix for dimensionality feature selection. The efficacy of the fire and smoke detection model is then compared with the following metrics such as: computational time, accuracy, precision, recall, and F1-score.