A number of different machine learning techniques are commonly used today. Different techniques are best suited to different problems. Some of the leading techniques are summarised below.
|Logistic Regression||A method of classifying elements in a data set into different categories by estimating probabilities using a distribution function.||Credit Scoring|
Measuring success of marketing campaigns
|Classification Tree||A hierarchy of binary classification questions used to classify input data. A set of trees is sometimes used in parallel to reduce the risk overspecialisation of a single tree.||Fraud detection|
Customer behaviour prediction
|Deep Learning||Multi-layered neural network, a network which imitates the function of a biological brain.||Computer vision|
Natural language processing
|Support Vector Machines||This methods seeks to find a mathematical function which correctly splits a set of data points into two categories.||Display advertising|
Image-based gender detection
Large-scale image classification
|Naive Bayes||A classification model based on application of Bayes’ theorem for a set of variables. It is ‘naive’ because it assumes that the variables are completely independent.||Email spam recognition|
News article classification
Automation Consultants has knowledge of all the above techniques and will employ the method best suited to your goals and your data.
Although the method is important, the quantity and quality of available data is perhaps even more so. Analysing and cleaning the data is a critical part of a successful ML project.