# Final Thoughts on Business Analytics Methods

08.02.2018

Episode #10 of the course Business analysis fundamentals by Polina Durneva

Good morning! Today, we will summarize all statistical and mathematical techniques that you have learned over the past nine days. I will also introduce you to a few other techniques that may spark your interest.

What Have We Learned?

K-Nearest Neighbor. We use the KNN algorithm when we want to classify a new record to a certain category using similarities and differences between this new record and other records in an existing dataset.

Naïve Bayes. Naïve Bayes uses conditional probability to classify a new record.

Association Rules. This method takes a form of, “if __, then ___,” and is mostly used to identify clusters of items that are bought together.

Cluster Analysis. Cluster Analysis is used to break a heterogeneous dataset into smaller chunks of homogeneous subsets.

Decision Tree Analysis. Decision Trees are used for classification and prediction. They take a form of flow chart that leads us to a certain solution to a problem.

Linear Regression. If we want to evaluate the impact of certain factors on something, we’ll stick with linear regression.

Logistic Regression. Similar in its form to linear regression, logistic regression is used to evaluate the probability that a new record belongs to a certain category.

Optimization. Optimization is used for maximization/minimization problems—for example, it helps us decide how we can maximize our profit by allocating different amounts of money to different departments.

Simulation. Simulation is a risk analysis that is based on uncertainty of future.

While we already covered the most popular techniques and methods used in business (and, of course, many other areas), there exist many other modeling techniques. Below is a summary of the most interesting ones.

Neural Nets. This is one of the most popular data-driven approaches in the field of artificial intelligence. Neural nets are used for classification and prediction. The algorithm behind this method tries to resemble the human brain. Basically, the algorithm tries to connect different chunks of information and derive something from them. To some extent, this method resembles linear and logistic regression in which you try to derive the effect on independent variables on the dependent variable. While when using regression, you decide what affects what, neural nets try to establish the cause-and-effect relationship on their own.

Discriminant Analysis. Discriminant analysis is a classification method. It is similar to cluster analysis, but it tries to predict the class of a new record by evaluating the distance between this new record and the average of the existing classes. Discriminant analysis provides us with the probability value.

Optimization in Simulation. The name of this method speaks for itself. We use it when we try to optimize something (maximize profit or minimize costs, for example) but are facing uncertainties in the future. Using this method, we can come up with the best decision variables (price or cost of a product) while accounting for possible changes in other parameters.

Of course, there exist many other methods in data mining, predictive analytics, and managerial science. I would also like to say that artificial intelligence plays a huge role in the development of new methods and iteration of the existing ones. Nowadays, we and our computers have access to big data, and many researchers are attempting to take the best advantage of this opportunity. Our entire humanity is moving toward automation and machine learning (a sub-field of artificial intelligence), and this could not be any more exciting. Over the course of ten days, we covered a variety of great methods used in machine learning, and I would encourage you not to stop learning about them in the future.

Take care,

Polina

Recommended book

Sprint: How to Solve Big Problems and Test New Ideas in Just Five Days Hardcover by Jake Knapp,‎ John Zeratsky,‎ Braden Kowitz

Share with friends