Details about Machine Learning
Before we start learning concepts about this machine learning tutorial, you could ask yourself why you need to learn machine learning? If you are from a business analyst or data analytics background, you might have questions like, how does analytics differ from machine learning (ML)? What is data mining, statistics or data science? I come from an analytics background, so I always had these questions until I learnt Machine Learning in depth.
These are the topics that will be covered in the machine learning tutorial.
- What is machine learning?
- Differences between all similar terminologies
- Why do we need machine learning?
- Tools used for machine learning
- Skills used for machine learning
- Future in machine learning
What is Machine Learning?
As the name suggests, a machine that is learning. Now we need to understand what exactly the machine learns? As we all know, a machine or computer can only understand 1 or 0’s (binary digits), so the machine only learns that for the given input data, what should be the output in the form of 1’s or 0’s (yes or no). Machines use complex algorithms to find patterns or correlation between the features and/or labels we provide into the input. In other terms, Machine Learning is nothing but an algorithm where it finds patterns within the given data and comes up with the best possible solution or outcome without any human involvement.
Differences between all similar terminologies
Let me explain to you the differences between multiple similar terms used in industry.
Artificial Intelligence (AI), as the name suggests, intelligence which was given artificially. AI gives computers the ability to replicate human intelligence. Machine learning is a subset of artificial intelligence.
Statistics is purely based on math, where it works on the hypothesis approach. It is more like that you know what will be the result of your experience or gut feeling, but you need to prove that mathematically. Statistics uses sample data to prove the data patterns assuming the sample is a true representation of the population. Unbalanced data in samples affects results in statistics. In contrast, machine learning is a process or algorithm where you let machines learn scientifically using the training data and predict the output from unseen data. ML doesn’t make any assumptions.
Data analytics is used to extract and transform data, understand the data, find trends/patterns, model building, visualization, and conclude. Data Analytics has four types: Descriptive Analytics, Diagnostic Analytics, Predictive Analytics and Prescriptive Analytics. Predictive Analytics interchangeably used with Machine Learning.
Predictive Analytics is nothing but an approach to creating predictive models, whereas Machine Learning is an algorithm. Predictive Analytics typically uses machine learning algorithms for predictive models.
Data Mining can be seen as a subset of data analytics. Data mining is used to find hidden patterns from the big structured dataset using statistical and mathematical models.
Data science is a very wide term that focuses not only on machine learning algorithms and statistics but also on entire data processing methodologies.
ML can learn patterns from structured (tabular data) and unstructured data (images, videos). We need complex algorithms (artificial neural networks), which mimics the structure of the human brain, to learn patterns from unstructured data. It is known as Deep Learning. Deep learning is a subset of machine learning.
You might have encountered that many teams do the analysis themselves and find patterns. They formulate those patterns and predict the output using those formulas. It is known as Rule-Based Machine Learning. These rules do not change or evolve with time until someone does the analysis again, unlike machine learning.
Why do we need Machine Learning?
As we move forward in this machine learning tutorial, let us understand why we need machine learning as we are in a big data world where data is easily available with minimal cost. It is very difficult to sustain in rule-based models, which are very costly to maintain and keep accurate with growing data. Machine learning is becoming the need of the time. It is fast and efficient. We can get pretty much real-time in Machine Learning.
Machine learning has many use cases now in almost all industries. Starting from automobile (tesla auto driverless cars), healthcare (cancer detection using certain features), marketing (Amazon’s recommendation for similar products), Social Media (face detection), smartphones (face unlock), banking (fraud detection) etc. It is an endless list that keeps growing with more research like GANs where you can create multiple images of one image.
Tools used for Machine Learning
There are three main tools which we use in machine learning:
- Databases (SQL, Redshift, S3, Oracle) – Databases and respective tools to pull data and pass to the algorithms
- Language (Python, R, SAS) – This is needed to code your algorithms to help machine learn the patterns
- Visualization (Tableau, Quicksight, PowerBI, excel) – This is needed to show your result to the end user
Skills used for Machine Learning
- Business understanding & structural thinking: It is necessary to have skill where you should be able to understand business problems, convert them into machine learning problems and provide solutions.
- Mathematics: You need to have knowledge of math (derivatives, logs etc.) to understand the algorithms and customize them as per your need.
- Software skills: You need this skill to write machine learning algorithms, deploy code into production and run it seamlessly.
Future in Machine Learning
Industries are evolving and transforming themselves to use more machine learning applications nowadays. The machine learning market is expected to grow USD 8.81 Billion by 2022, at a Compound Annual Growth Rate (CAGR) of 44.1% during the forecast period.
This brings us to the end of the machine learning tutorial. Thanks so much for staying with me so far. Happy Learning!