Ensemble Machine Learning Technique: Voting

In this video, we're specifically focusing on the concept of voting under parallel ensembling. 🌟 Mindmap Refresh: Before we delve into voting, let's refresh our minds with the ensembling mindmap. Ensembling techniques can be broadly categorized into three types: sequential ensembling, parallel ensembling, and composite ensembling. Voting falls under the parallel ensembling category, where multiple base models are trained independently. 🧠🔄 Voting vs. Bagging: One key difference between voting and bagging is how the final prediction is made. In bagging, all base models are of same type and have equal weights, and the final prediction is based on averaging or majority voting. In contrast, voting allows for assigning different weights to base models and offers hard and soft voting options. 🤝🎒 Assigning Weights: Assigning weights to different base models is a critical aspect of voting. These weights can be based on various factors such as model performance, confidence levels, or domain expertise. By assigning weights, we can give more importance to the predictions of certain models over others. ⚖️🔢 Hard vs. Soft Voting: In hard voting, the final prediction is based on the majority vote of the base models. Each model gets one vote, and the class with the majority of votes is chosen. On the other hand, soft voting considers the predicted probabilities of each class from each model. The class with the highest sum of probabilities is selected as the final prediction. 🗳️🔢 Conclusion: In conclusion, voting in parallel ensembling is a powerful technique that allows us to combine the predictions of multiple base models to make more accurate and robust predictions. By assigning weights and choosing between hard and soft voting, we can tailor the voting strategy to suit our specific needs and improve the overall performance of our ensemble models. 🚀🎓