Machine learning: What really is machine learning?
Machine Learning is a way of writing codes to teach computers to do what humans and animals naturally do like learning from experience. Machine Learning enables computer to reason from pass experience and other factors which human reason with .
Machine Learning algorithm use computational mathematical methods to learn information from data without relying on a predetermined equation as a model. The algorithm adaptively improves their performance as the number of samples available for learning increases.
Machine learning algorithms find natural patterns in data that generates insight and help you make better decisions and predictions. They are used everyday to make critical decisions in medical diagnosis, stock trading, energy load forecasting, facial recognition and more. Media sites require machine learning to sift through millions of options to give you song or movie recommendations which you might actually need at that moment. Retailers use it to gain insight into their customers, purchasing behavior.
Real world applications of Machine Learning
With rise in big data, machine learning has become particularly important for solving problems in areas like this:
- Computational finance, for credit and storing algorithmic trading
- Image processing and computer vision for face recognition, motion detection and object detection
- Computational biology for tumor detection, drug discovery and DNA sequencing
- Energy production for price and load fore casting
- Automotive, aerospace and manufacturing for predictive maintenance
- Natural language processing
Segment in Machine Learning
Machine learning uses three types of techniques:
1. Supervised Learning
These sets of algorithm train a model on known input and output data so that it can predict future outputs. The main objective of supervised learning is to build a model that makes prediction based on evidence in the presence of uncertainty. Supervised Learning takes set of input data and trains a model to make prediction for new set of data. Supervised Learning uses the classification and regression technique to develop predictive model.
This technique predicts response classifying input data into categories . Some of its applications are medical imaging, Speech recognition and credit scoring.
The algorithms in the classification technique are Support Vector Machine, Discriminant Analysis, Naïve Bayes and Nearest Neighbor.
This technique predicts continuous responses. Some of its applications are electricity load forecasting and algorithmic trading.
The algorithms in the regression technique are Linear Regression, GLM, SVR, GPR, Ensemble Methods, Decision Trees and Neural Networks.
2. Unsupervised Learning
These sets of algorithm train a model by finding patterns or intrinsic structures in input data. It is used to get features from dataset and label the data according to their features. Unsupervised learning uses the clustering and association technique.
This is the most common technique of the unsupervised learning. Clustering explores datasets to find hidden patterns or groupings in data. Some of its applications are gene sequence analysis, Market research and object recognition.
The algorithms in the clustering are K-Means, K-Medoids, Fuzzy, C-Means, Hierarchical, Gaussian Mixture, Neural Networks and Hidden Markov Model.
This technique predicts using the associate of the dataset. It finds the features and predicts the associates. Some of its applications are predicting what a customer would buy if he buys a products from a particular store online.
3 Re-enforcement Learning
These set of algorithm trains a model to teach itself based on the past action reward.
The algorithms in the re-enforcement learning are Q-learning, Temporal Difference and Deep Adversarial Networks.
Machine Learning Workflow
There are some patterns to solve a particular problem with machine learning. Some of which are;
• Define the objective
This involves the stating the problem and analyzing what you want to use to solve it and how you want to solve it. Before collection of data, the objectives are stated down. This shows the type of data to be collected, where the data would be collected and how the data would be collected from the source.
• Collect the data
After the definition of the objectives, the right or necessary data are collected from the source. Collection of data can take a lot of time depending on the soure.
• Preprocess the data
Since raw data are not usually clean, it is necessary to clean the data so as to improve the usage of the data. The features and structure of the data is also checked for more information on which category of algorithm to use.
• Set algorithm
After cleaning the data, an algorithm would be selected to train the model. This might take time depending on the skillfulness of the data scientist. The reason being is because the first algorithm might not be the right match up for the data. The data scientist chooses an algorithm.
• Train the model
After selecting the algorithm, the data are now fed to the algorithm to train the model and improve accuracy a lot of data is used. Training a model allows the computer or machine to understand the data
• Test the model
After training the model, the model is then tested with other new data and if the desired or accurate result is achieved. It would then be moved to the next stage if not, it would be taken back to the stage in which a new algorithm is required.l
After, all is set prediction takes place.
Check my article on what is programming