EP 4. MACHINE LEARNING CYCLE

EP 4. MACHINE LEARNING CYCLE:

Creating a machine learning application or implementing a machine learning algorithm to some data is an iterative process . We can't simply develop a machine learning algorithm and just leave it because data changes , preferences emerge. When we train our machine learning model with a machine learning algorithm, for each and every algorithm there is different accuracy. So we should choose which one fits correctly in terms of all metrics like f-square, mean absolute error, p-value etc. The machine learning cycle is continous cycle  and choosing which algorithm fits is only one of the step in the cycle.The steps in the machine learning cycle are as follows:

1> Identify the data: First of all when we have a dataset full of data and features(columns) we have to select which feature we want because it will be an underfit when we try to predict. We can do feature engineering when we want to add any features.Then the next step is to observe how the data is like whether it is numerical data or categorical data. So based on the type of data we can remove some rows which are not efficient for our model.

2>Prepare the data: We have to make sure that our data is clean, secured. If we create a machine learning algorithm with inaccurate data then our model will have less accuracy or the complete application may fail.

3>Select the machine learning algorithm: We must select a machine learning algorithm that actually suits our model to do our desired work.

4>Train: We need to train the algorithm depending upon the type of data and algorithm, the training data may be Supervised, Unsupervised, or Reinforcement learning.

5>Predict:  After deployment , start making predictions based on new,incoming data and we also get to choose whether to accept the model or reject the model based on the accuracy,  mean absolute error and the other metrics present.