Data Mining and Machine Learning

In this article, you’ll learn more about data mining and machine learning.

Data Mining

Everyone keeps talking about data mining being a big deal in the current data analysis space; it indeed is. Let us see how.

The M&S Story

Marks & Spencer, fondly known as M&S, is a favourite and trusted retailer in Britain. In the current state of competition, no such ‘reputed’ or ‘trusted’ company can solely rely on their reputation. Hence, planning to adopt a data-driven approach to their business was one of their best decisions.

M&S today relies on real data from internal sources such as charge card systems, and external sources such as census, demographic, and national panel data. In addition, through cluster analysis and data mining, M&S succeeded to identify 11 core customer segments. [1]

Previously, they used to stock their stores based on the number of customers per square footage. However, today, leveraging data mining techniques, careful customer profile analysis, and behaviour analysis is conducted to identify the ideal store stock quantities.

What Does ‘Data Mining’ Mean?

The term ‘data mining’ is a misnomer, because it is primarily concerned with discovering patterns and anomalies within data sets, but is not related to extraction of the data itself. It is the process of uncovering patterns and finding anomalies and relationships in large data sets that can be used to make predictions about future trends.

Practical Applications

Here are some real-life applications of data mining:

Database marketing and targeting: Retailers use data mining to understand their customers better by segmenting market groups, and by tailoring promotions that drill down to offer customised promotions.

Fraud detection and preventions: Financial institutions used to automatically detect and stop fraudulent transactions by tracking the behind-the-scenes activity of each transaction – sometimes even without the consumer’s knowledge!

Credit risk management: Banks deploy data mining models to predict a borrower’s ability to take on and repay debt. Using a variety of demographic and personal information, these models automatically select an interest rate based on the level of risk assigned to the client.

Spam filtering: Systems can analyse the common characteristics of millions of malicious messages to inform the development of security software.

Machine Learning

Machine learning is a vast field in itself and there are various nuances and complexities associated with each of these algorithms. In this course, you are not going to cover the machine learning algorithms and its implementation, rather the purpose is to introduce you to understand its positioning in the overall data analytics field.

What Does ‘Machine Learning’ Mean?

Machine learning is the data analytics technique that teaches computers to learn from experiences; that is, learn from data sets that have known results. This is achieved using computational techniques rooted in statistical analysis of data; using these techniques, the computer iteratively learns through the input data and it is able to perform above a certain expected threshold.

Once the machine learning algorithm has been trained, it can then be applied to unknown data sets to make inferences.

Practical Applications

There are broadly two types of machine learning approaches:

Supervised machine learning (algorithms): In this, algorithms are the ones that are trained on the data set that has label output data, sometimes also referred to as ground truth.

Unsupervised machine learning (algorithms): In this, algorithms are trained to understand data with labelled output, without any ground truth.

Examples of Supervised Machine Learning

Linear regressions: Linear regressions predict the value of a continuous variable using one or more independent input variables.

For example, houses based on area, frontage, year built, zip code, and so on.

Logistic regressions: Logistic regressions predict the probability of a categorical variable using one or more independent inputs.

For example, banks use logistics regression to predict the probability of whether a loan applicant will default or not based on credit score, household income, dependents, and other personal factors.

Classification or regression decision trees: Decision trees is a predictive modelling technique in which a model creates sets of binary rules to split and group the highest proportion of similar target variables together iteratively over a specified depth.

For example, a model built to categorise health insurance policies to identify mortality risk.

Random forest: Random forest is where multiple decision trees are augmented together to collectively make a prediction by averaging the results of all the multiple decision trees.

Examples of Unsupervised Machine Learning

Clustering algorithm: Clustering algorithms are used to categorise and group different data points or sets. If an analyst wants to discover interesting data points that might add value to a pattern, they can group them using this technique.

For example, identifying fake news, filtering spam emails, or classifying the network of traffic.

Association analysis: Association analysis is where one has to identify and capture associations between various items and events. When you chase these associations, you tend to uncover intriguing relationships.

For example, the famous beer-diaper association discovered by a renowned supermarket.

Principal component analysis: Principal component analysis deals more closely with summarising the data.

Think about it

Information is the new gold, and we can extract that from our data. However, data comes in all sizes!

• Do you think there are any challenges between mining huge amounts of data as compared to mining smaller amounts of data?
• How is data mining different from machine learning in your organisation?

Share your rationale with an example or scenario eliciting the difference or similarities in the comments.

References

_{1. Giraud-Carrier C. Success Stories in Data/Text Mining [Internet]. Brigham Young University; [date unknown]. Available from: http://dml.cs.byu.edu/~cgc/docs/mldm_tools/Reading/DMSuccessStories.html#_ftnref}