About

What is ML?

Machine Learning (ML) is defined as follows: A code learns from experience E with respect to a task T and a performance measure P, if its performance on T, as measured by P, improves with E. [1].

Example 1: your code monitors spam and classifies emails as spam or not spam (SNS). In this case, T = classifying emails as SNS; E = watching you label emails as SNS; P = the fraction of emails correctly classified as SNS.

Example 2: playing checkers. E = the experience of playing many games of checkers; T = the task of playing checkers; P = the probability that the program will win the next game.

ML is a part of Artificial Intelligence (AI). ML algorithms build a model based on sample data, known as training data, in order to make predictions or decisions without being explicitly programmed to do so [2]. ML is an important subset of data science. Through the use of statistical methods, data science algorithms are trained to make classifications or predictions, uncovering key insights within data mining projects. These insights subsequently drive decision making within applications and businesses, ideally impacting key BI/fintech metrics.

Bottom Line: ML is the science of getting computers to learn, without being explicitly programmed.

Why ML?

Modern businesses across almost all sectors use various machine learning technologies [4]: logistics, manufacturing, hospitality, travel/tourism, energy, and utilities. As an example, the healthcare industry is utilizing ML to achieve more accurate diagnoses. Retailers also use ML to send the right goods and products to the right stores before it is out of stock. The top ML applications are chatbots as virtual assistants or VA, BI-type decision support, customer recommendation and churn engines, demand pricing, market research, fraud detection, computer vision (CV), operational efficiencies via DevSecOps known as MLOps, and text data analysis via natural language processing (NLP). Very successful ML use cases deal with database mining by analyzing large datasets such as web click data, medical records, etc. Industry-4 IoT devices can also perform ML tasks. Today most innovation-oriented companies are looking for ways to make their brand stand out in the marketplace.

ML Tech/Ops

Your ML journey begins with the most fundamental question:

Here DA means Data Analytics that consists of descriptive, diagnostic (such as Hypothesis, Conjectures and Assumptions), predictive and prescriptive algorithms [5]. Descriptive DA includes Trends, Averages, STD, Box Plots, X-Plots, Hist, etc. Predictive DA is based upon the Regression Analysis and Logistical Regression. Prescriptive DA is designed for Decision Making (Concept of Probability).

In contrast, ML implies unsupervised, supervised, reinforcement and deep learning algorithms [3]. Unsupervised ML deals with Untrained Data Mining

(Classification, Regression and K-means Clustering as implemented in scikit-learn [6]). With Anaconda’s platform [7], you can build and deploy deep learning (DL) algorithms that use neural networks. Anaconda easily integrates with tools like TensorFlow, pyTorch, mxnet and Keras [8] so you can build and train neural network models, including convolutional neural networks (CNNs) and generative adversarial networks (GANs). Reinforcement learning (RL) such as Q-learning or SARSA is an area of ML concerned with how intelligent agents ought to take actions in an environment in order to maximize the notion of cumulative reward [9]. The free Mathworks RL onramp [10] provides a vivid introduction to RL for control problems.

Key Takeaways

ML: parse data, learn from that data, and make informed decisions based on what it has learned.

DL: create NN layers that can learn and make intelligent decisions on its own (e.g. via backpropagation loop).

In fact, most SMEs these days use public cloud computing web services to use ML for a fee so that they can focus on their core business rather than building on-premise ML infrastructure. The AWS ML API’s such as Sagemaker and Rekognition enable data scientists and MLOps engineers to build, train, test and deploy ML models for various use cases [11]. The Azure ML Studio accelerates time to value with industry-leading MLOps, open-source interoperability, and integrated tools such as DevOps Boards [12]. In addition, you can use the GCP AI Platform to train your ML models at scale, to host your trained model in the cloud, and to use your model to make predictions about new data [13].

ML Use Cases

AWS ML services [11] explore a variety of common use cases such as intelligent contact centers, personalization, automated document processing, search Kendra, fraud detection, business metrics and media content analysis. The Azure ML is helping retail and consumer brands improve the shopping experience by ensuring shelves are stocked and product is always available when, where and how the consumer wants to shop [12]. The top three GCP ML use cases are as follows: predicting high value start-up investment, predicting maintenance jobs on an oil rig, and real-time predicted relevant ads for customers in taxis.

References

[1] Mitchell, T., 1997, Machine learning, McGraw Hill.

[2] https://en.wikipedia.org/wiki/Machine_learning

[3] https://www.ibm.com/cloud/learn/machine-learning

[4] https://www.techfunnel.com/information-technology/machine-learning-business-applications/

[5] https://www.investopedia.com/terms/d/data-analytics.asp

[6] https://scikit-learn.org/stable/

[7] https://www.anaconda.com/use-cases

[8] https://towardsdatascience.com/best-python-libraries-for-machine-learning-and-deep-learning

[9] https://en.wikipedia.org/wiki/Reinforcement_learning

[10] https://www.mathworks.com/learn/tutorials/reinforcement-learning-onramp.html

[11] https://aws.amazon.com/machine-learning/