As explained in my previous post in “Introduction to Machine Learning”.
After defining the problem, we have to collect the data which is relevant to the problem.
We have 2 kinds of data.
1. Seen/ Trained Data: We define the problem when the class labels are known
2. Unseen/ Test Data: Whatever the Machine Learning model we test on the data and see the results, here we don’t know what data has and class labels are unknown.
How a data is interpreted?
In a data set,
· Each row/ a data point is called as Instance.
· Each column is called as Feature.
· The nth label / last label is called Outlook.
Based on the data set we try to predict the outcome. We have different types of predictions:
· When we predict a numerical value then it is Regression.
· When we predict from a category then it is classification. (Regression predicts a real time value)
We normally divide the available dataset in 80–20. 80% of data we train the algorithms and by implementing the multiple Iterations, to find the best possible output which has minimal number of errors and 20% of the data we use it as test data by using the algorithm which is the outcome of trained dataset.
Errors are minimised by applying the optimisation algorithm which is called Gradient Decent.
So, now we can see different types of Machine Learning catergories:
1. Supervise Learning: In simple terms, we are given features along with the labels. It can be class labels/ task labels.
2. Unsupervised Learning: Only features are given
3. Reinforcement Learning: It is something doesn’t know anything it just does which is being inputted and projects in a same way. So, it acts how we want like a baby.
Machine learning Algorithms:
“More data, better model you get”- to get effective ML model.
Further I would like to learn about Machine learning and share the information. Thank you for reading.