6 Steps of Machine Learning

• Six Steps of Machine Learning -

  -> Data Collection
  -> Data Cleaning
  -> Data Analysis
  -> Machine Learning Models
  -> Analysis
  -> Iterate
  

Why Machine Learning?

Machine learning is basically the study of statistical computer algorithms that improve automatically through the data provided. It is basically teaching machines to teach themselves.
We live in a world where everything around us can provide valuable information to learn, adapt, and alter. If we can use these data points around us and tune them for our needs accordingly, they will be more useful.
We live in a world that is running past our pace. As technology grows, so do we.

-Technicality-

The data collected from various sources and methods can be analyzed, extracted useful pieces of information, and later, trained on those pieces of information on several machine learning models.

How does Data differ from Information?

Information>>Data

Anything that you collect is essentially data. These useful pieces of text, images, and videos, can be called information. Information is essentially extracted from data.

-Data Collection And Analysis-

You now understand the importance of data and the need to use it in an effective way.
Now, let’s dive deeper into the art of data collection and analysis.
Data collection is mainly done by using web scraping techniques. It is later used for further cleaning and analyzing processes.
Collected data from various sources can have numerous useful information in it.

Data cleaning is preparing the data for further analysis by removing duplicate values, fixing missing values, unformatted variables, irrelevant information, etc.

Ironically, the majority of the time is spent on this process in the entire ML workflow. It is a crucial step in the workflow of Machine Learning, as negligence in this step can cause large errors and problems in the machine learning model.

So far, we have covered the first 3 steps of Machine Learning.

-Machine Learning Models-

Machine Learning Models can be classified into 4.

• Supervised learning

• Semi-Supervised learning

• Unsupervised learning

• Reinforcement learning

In this context, we are not going to dive deeper into the details of each model.

Supervised learning -> Subset of Machine Learning, which uses labeled data for its predictions and classification-based outputs.
Unsupervised learning -> Machine Learning model, which does not use labeled data and uses machine learning algorithms for its analysis and clustered outputs. These models learn patterns from the unlabeled data.

Semi-supervised learning -> Machine Learning model, which includes supervised and unsupervised learning simultaneously. It falls in between supervised and unsupervised learning models. The model trains on a small amount of labeled data and a large amount of unlabeled data.

Reinforcement learning -> Machine Learning model, which helps in the decision-making. For a suitable/favorable outcome, it gives rewards, and unfavorable outcome, it punishes. It is the science of outputting optimal outcomes in a given environment. Unlike the supervised learning model, in the reinforcement learning model, there is no answer key for training the model. The model learns from its previous prediction.

-Analysis-

To measure/evaluate the performance of the prediction of a particular model, we need performance metrics.
For example, the most common evaluation metrics designed for regression tasks are,
· Mean Absolute Error (Mean of absolute value of errors)
· Mean Squared Error (Larger errors are punished here more than MAE)
· Root Mean Squared Error (Root of the MSE, outputs in the unit of label)

-Iterate-

Last but not least, “Iterate”. Do what you have done previously again by making necessary changes until you get the desired outcome/performance. You can change, say, for example, change hyperparameters for better results from the model.

CONCLUSION

In this short introduction to Machine Learning, I hope you got a basic idea of what machine learning workflow looks like. However, there are a lot of other steps as we get deeper. But in short, I think, this is it. We made our surface clean!

Thank You

Written: Sreehari J R
LinkedIn: Click Here