ML analytic models to consider

Machine learning models can be applied to various types of analytics, including descriptive, predictive, and prescriptive analytics.

Here's how machine learning models are used in each of these analytics domains:

​ 1. Descriptive Analytics:

• Definition: Descriptive analytics involves summarizing historical data to understand what has happened in the past. It provides insights into patterns, trends, and key performance indicators (KPIs) to describe the current state of affairs.

• Machine Learning Models: In descriptive analytics, machine learning models are not typically the primary focus. Instead, basic statistical and visualization techniques are often used to explore and present historical data. Tools like histograms, scatter plots, and summary statistics can be applied to understand data distributions, correlations, and anomalies.

2. Predictive Analytics:

• Definition: Predictive analytics is concerned with forecasting future events or outcomes based on historical data. It aims to answer questions like "What is likely to happen next?" by building predictive models.

• Machine Learning Models: Various machine learning models are commonly used in predictive analytics, including:

• Linear Regression: It predicts a continuous target variable based on one or more input features.

• Logistic Regression: Used for binary classification tasks where the goal is to predict one of two classes.

• Decision Trees and Random Forests: These models are suitable for both classification and regression tasks and provide insights into feature importance.

• Support Vector Machines (SVM): Useful for classification and regression, SVMs aim to find a hyperplane that best separates data points.

• Neural Networks: Deep learning models, including feedforward and recurrent neural networks, are used for complex prediction tasks.

• Time Series Models: Models like ARIMA (AutoRegressive Integrated Moving Average) are used for time series forecasting.

• Gradient Boosting Machines: Algorithms like XGBoost and LightGBM are powerful for various prediction tasks.

​3. Prescriptive Analytics:

• Definition: Prescriptive analytics goes beyond predicting outcomes and offers recommendations or actions to optimize decision-making. It helps answer questions like "What should we do to achieve a desired outcome?"

• Machine Learning Models: Machine learning models can be used to provide recommendations and insights in prescriptive analytics. Some of the approaches include:

• Recommender Systems: These systems, like collaborative filtering or content-based filtering, suggest products, content, or actions based on user behavior and preferences.

• Optimization Algorithms: Linear programming, mixed-integer programming, and other optimization techniques are used to find the best solutions to complex problems, such as resource allocation or scheduling.

• Simulation Models: Simulation-based models allow organizations to simulate different scenarios and assess the impact of various decisions before implementing them in the real world.

• Constraint-based Models: These models take into account business rules and constraints to generate recommendations that adhere to specific guidelines.

• A/B Testing and Experimentation: While not traditional machine learning models, A/B testing and experimentation are often used to test different strategies and make data-driven decisions.

In general, descriptive analytics uses basic statistics and visualizations, predictive analytics leverages a wide range of machine learning models for forecasting, and prescriptive analytics employs models that recommend actions and decisions to optimize outcomes.

Sash Barige

Mar/20/2022

Further Read:

Descriptive Analytics:

• Overview article from Tableau: https://www.tableau.com/learn/articles/descriptive-analytics

• Types of descriptive analytics from Oracle: https://www.oracle.com/business-analytics/what-is-descriptive-analytics/

• Example use cases from Qualtrics: https://www.qualtrics.com/experience-management/customer/descriptive-analytics/

Predictive Analytics:

• Introduction from SAS: https://www.sas.com/en_us/insights/analytics/predictive-analytics.html

• Real world examples from FICO: https://www.fico.com/blogs/4-examples-real-world-predictive-analytics

• Overview video from IBM: https://www.youtube.com/watch?v=vWtUHh2kjRs

Prescriptive Analytics:

• Explanation from Deloitte: https://www2.deloitte.com/us/en/pages/deloitte-analytics/articles/prescriptive-analytics.html

• Use case examples from Gartner: https://www.gartner.com/smarterwithgartner/how-to-get-started-with-prescriptive-analytics/

• Intro article from Harvard Business Review: https://hbr.org/2019/03/prescriptive-analytics-the-ultimate-how-to-guide