Machine Learning: Starter
A Quick History of Machine Learning!
What is Machine Learning?
- The goal from the start had been to create a machine that could think and learn like a person.
- Without being explicitly programmed, machine learning techniques allow computers to learn from data and even improve themselves.
What is Machine Learning in simple words?
When you combine computer programming skills with Statistics, the result is a field of study known as Machine Learning.
When Do We Use Machine Learning?
Machine learning is shown to improve and bring change in a variety of areas. It doesn’t matter what industry it is, from farming to spaceflight.
ML Use-cases:
- Human expertise does not exist (navigating on Mars)
- Humans can’t explain their expertise (speech recognition)
- Models must be customized (personalized medicine)
- Models are based on huge amounts of data (genomics)
How does Machine Learning work?
Steps In Machine Learning
There are four steps:
- Data Preparation: The process of cleaning and altering raw data before processing and analysis are known as data preparation. It’s a crucial stage before processing that often include reformatting data, making data changes, and integrating data sets to enrich data.
- Feature Engineering: The act of selecting, altering, and transforming raw data into features that can be used in supervised learning is known as feature engineering. It may be necessary to build and train better features for machine learning to perform well on new jobs.
- Data Modeling: Data modeling is the practice of using words and symbols to represent data and how it flows in a simplified representation of a software system and the data pieces it contains. Data models serve as a roadmap for creating a new database or reengineering an existing one.
- Performance: The performance of a machine learning model is relative, and thoughts about what score a good model can attain only make sense and can be read in the context of the skill scores of other models trained on the same data.
What are the different types of machine learning?
There are three major types of machine learning:
- Supervised ML: Exercising the algorithm is a good way to learn it. The output should be y if the input is X. (Target Variable). Linear regression, logistic regression, decision trees, random forests, SVM, naive Bayes, XGboost, AdaBoost, and other common methods are used for this.
- Unsupervised ML: There is no Target variable. In other words, you don’t have data indicating that the input is X and the outcome is y. Important patterns can be extracted immediately from data in this case, such as clustering similar sorts of rows (Clustering), without any prior knowledge of the data. K-Means, DBSCAN, PCA, ICA, and Apriori are some common algorithms.
- Reinforcement ML: When a Machine Learning algorithm learns from its failures and improves to attain a goal in the next iteration. Monte Carlo, Q-Learning, and SARSA are three common algorithms.
What are different types of Supervised machine learning?
There are two different types of Supervised Machine Learning Algorithms:
- Regression
- Classification
- Regression: When predicting a continuous number such as “Sales,” “Profit,” “Demand,” “Turnover,” “Volumes,” “Number of Visitors,” and so on, you use supervised machine learning. This is the concept of regression.
- Classification: When you are predicting a categorical/Discrete value like “0/1”, “Yes/No”, “Good/Bad”, “Silver/Gold/Platinum” etc. This is Classification.
- There are various algorithms used in supervised machine learning, and I’ve listed a few of the most prominent ones here.
What are the different types of Unsupervised machine learning?
There are three different types of Unsupervised Machine Learning Algorithms:
- Clustering
- Dimensionality Reduction
- Association
- Clustering: Putting together groupings of related rows. The objective is to group all of the data rows that are of the same type into a single cluster/Group. K-Means, Hierarchical clustering, DBSCAN, and OPTICS are all useful clustering techniques.
- Dimension Reduction: By consolidating comparable predictor columns into a single column, you can reduce the number of predictor columns in your data. Dimension reduction is the term for this procedure. Because each dimension in your data is represented by a column, The efficiency of predictive models in terms of speed and accuracy is reduced when the data is high dimensional (for example, 800 columns). As a result, it would be fantastic if we could reduce those 800 columns to 10–15 columns that represent the same patterns as all 800 columns. PCA, ICA, T-SNA, and UMAP are some of the Algorithms used for Dimension Reduction.
- Association: The most common use of Association rule mining is to figure out which goods sell well together. To understand associations, rules are developed by counting transaction items and calculating support, confidence, and lift scores. In basic terms, how probable is it that the user will buy item B if they buy item A? Apriori, Eclat, and FP-Growth are the algorithms used to find associations.
What are different types of Reinforcement learning?
- Reinforcement learning is the type of machine learning where the algorithms learn by making mistakes and improving in the next iteration.
- The main idea is to give a reward (one point) if the algorithm takes the correct step, and a punishment (one point) if the program does the wrong step. It’s almost as if you’re training a youngster! However, you do not oversee the learning in this case; instead, you merely describe the rewards and punishments and then leave the algorithm to work on its own, receiving feedback.
Some of the algorithms used for reinforcement learning are:
- Monte Carlo
- Q-Learning
- SARSA (State-Action-Reward-State-Action)
The preceding is an overview of the fundamentals of Machine Learning that should be understood before going deeper. I’ll continue to write articles that will gradually teach us about the many ideas of machine learning.
Hope this helps :) Follow if you like my posts. Let’s connect on LinkedIn.
Happy learning 😃
