Sometimes, one machine learning model is not good enough for a certain scenario or use case as it might not give you the desired accuracy, recall, and precision. Hence, multiple learning models—or an ensemble of models captures the pattern of the data and gives better output.
As an example, let's say we are trying to decide on a place where we would like to go in the summer. Typically, if we are planning for a trip, the suggestions for the place pours in from all corners. That is, these suggestions might come from our family, websites, friends, and travel agencies, and then we have to decide on the basis of a good experience that we had in the past:
- Family: Let's say that whenever we have consulted a family member and listened to them, there has been a 60% chance that they were proven right and we ended up having a good experience on the trip.
- Friends: Similarly, if we listen to our friends, they suggest places where we might have a good experience. In these instances, a good experience occurred in 50% of cases.
- Travel websites: Travel websites are another source where we can get loads of information regarding where to visit. If we choose to take their advice, there's a 35% chance that they were right and we had a good experience.
- Travel agencies: Another piece of advice and information might flow from travel agencies if we go and check with them first. Based on our past experiences, we saw that they were right in 45% of cases.
However, we have to accumulate all of the preceding inputs and make a decision since no source has been 100% correct so far. If we combine these results, the accuracy scenario will be as follows:
1 - (60% * 50% * 35% * 45%) 1- 0.04725 = 0.95275
# Accuracy is close to 95%.
From this, we are able to see the impact of ensemble modeling.