predictive modeling - xl your mind · models are at the heart of predictive analytics. ... gearing...
TRANSCRIPT
Predictive Modeling Improve your decision-making using a variety of modeling techniques. Apply your old
data to answer new questions.
Duration:
Two days (16 hours)
Contact Us: +49 211 99 54 89 84 [email protected]
What to Expect? This two-day training explores predictive modeling as a decision-making aid. We will discuss types of models, modeling process and performance measurements. This training covers a wide range of modeling techniques such as regression analysis, machine learning, cluster analysis, ensemble methods.
This training will enable you to analyze your data with sophisticated statistical techniques and use your findings to gain valuable insights about your business.
Objectives
This training is designed for analysts who want to use models to answer business questions or employ machine learning. Prior knowledge of basic Excel functionalities and statistics is beneficial.
Who is It For? Is i
About us The transformation of data into meaningful information is crucial for making the right business decisions. We help you to manage the ever widening stream of data in the most time-efficient ways. With trainings that are adjusted to your industry and function, you will quickly become an expert in organizing, analyzing, and visualizing data to gain more insights. In a nutshell - our trainings enable you to do more work in significantly less time and with better results.
Training Topics
Models are at the heart of predictive analytics. They help us make sound business decisions. The training starts with clarifying central notions behind models, their components, and the common distinctions used to categorize them. Each step of the predictive modeling process will be presented in order to reach the desired outcome - from model training and assessment to deployment. Example questions What is the difference between supervised and unsupervised models? How does the model building process look like and what does each stage involve? What are the core components of a model? Focus points
Types of models
Understanding the difference between various model classifications.
The process of predictive modeling Looking at each stage of the modeling process supports the desired outcome and will help the decision making.
Performance metrics Delving into measures of performance in the context of describing the accuracy of different models.
Regression models determine how variables are related and enable you to predict the behavior of one variable using knowledge of other variables. They often constitute a first step into predictive modeling as they are relatively easy to train and interpret. Regression analysis also covers quantification of uncertainty around the model’s prediction by creating prediction intervals. Application example Estimating the expected claim costs for a specific client segment.
Gearing Up for Predictive Modeling
Regression Models
Example questions
How to pick the right variables for a regression model? How to do a regression analysis with the Analysis Toolpak, Solver, and functions? How to perform validity checks for a regression model? Focus points Single and multivariate linear regression
Training a linear regression model by minimizing the sum of squared errors with Solver and the regression Analysis Toolpak.
Assessing linear regression models Evaluating model performance on a holdout set and calculating the significance of coefficients.
Dummy variables, nonlinearity and interaction Improving a model by accounting for nonlinear effects.
Many decisions deal with understanding or estimating probabilities associated with certain events or behaviors. Frequently, these events or behaviors tend to have two possible outcomes like when predicting if a client will leave the company by the end of the year or not. Binomial outcomes present special problems that standard linear regression does not handle well. This part of the training shows how they can be analyzed using logistic regression models instead. Application example Estimating the likelihood that a client will file an insurance claim. Example questions
Why is logistic regression necessary? What role does maximum likelihood estimation play in a logistic regression? How to interpret logistic regression coefficients? Focus points
Logistic link model Adding a logistic link function to a general linear model and reoptimizing.
Model calibration Maximizing likelihood in a logistic regression using Solver.
Logistic Regression
Assessing logistic regression models
Comparing models with the ROC curve.
Neural networks are an amazing form of artificial intelligence that can capture complex relationships. Essentially a neural network is a “black box” that searches many models to find a relationship involving independent variables that can predict the value of a dependent variable most accurately. The usage of neural networks is increasing rapidly as they have been successfully applied in many situations. You will learn how to apply neural networks and why this technique plays a key role in the field of analytics. Application example Forecasting insurance premiums revenue. Example questions
What are neural networks and how do they help solve regression and classification problems? How to decide on the number of layers and the number of processing elements per layer? What variables are used in neural networks that support self-driving cars or facial recognition? Focus points
Training the network Exploring the iterative learning process of neural networks.
Testing for accuracy Understanding how to achieve better test accuracy with neural networks.
Making predictions
Creating a forecast for new data and evaluating accuracy.
Ensemble methods combine several trained models in order to produce a single model that is more powerful than each one individually. The training introduces bootstrap aggregation (bagging) and boosting, currently the two most popular modeling techniques used in business.
Neural Networks
Ensemble Methods
Application example Fraud prediction to help insurance companies optimally utilize resources and reduce fraud losses. Example questions
What is an ensemble model and what are its benefits? What are bagging and boosting? How to construct an overall prediction and evaluate the model’s performance? Focus points
Bagging Bagging is one ensemble technique that uses different samples of the same data to create multiple versions of the same model.
Boosting Boosting takes the whole dataset, but assigns different weights to observations in order to end up with a boosted model.
Cluster analysis helps to tease out relationships in datasets when you do not know what the key dimensions are. It is; therefore, an ideal technique for exploration. It helps to categorize objects into groups in a way that objects in each group are substantially different from objects in the other groups. Application example Classification of policy holders according to their perceived risk. Example questions
How to measure similarity between observations? How to decide on the number of clusters? How to create simple classification rules to be used by people without statistical training? Focus points
Standardizing attributes Making analysis unit-free so that each attribute has the same effect on the cluster selection.
Cluster Analysis
Distance metrics Choosing the appropriate distance measure for a problem set.
Solver model for cluster analysis Finding optimal clusters using Solver and interpreting the results.
Business Focus
You will learn practical skills and methods that can be applied straight away. Exercises and cases are based on common business challenges and the content is continuously validated by working professionals from different industries.
Continuing Support
To ensure that you can apply what you have learned during our trainings, our team stands
ready to give you a timely response to any questions you still might have.
Flexible
We take greatest care that the training content as well as facilities match your training
needs
International
We operate internationally and the training content as well as delivery method are very
well suited to professionals working in multinational organizations.
Why Choose our Trainings?
International
We operate internationally and the training content as well as delivery method are very
well suited to professionals working in multinational organizations.
Expert Knowledge
Our trainers are experts in the field of data analysis and automation with Excel. They have
helped many companies build cost effective solutions with Excel and incorporate this
experience in our trainings.
Continuing Support
To ensure that you can apply what you have learned during our trainings, our team stands
ready to give you a timely response to any questions you may still have.
Flexible
We take greatest care that the training content as well as the facilities satisfy your training
requirements.
Business Focus
You will learn practical skills and methods that can be applied straight away. Exercises and
cases are based on common business challenges while content is continuously validated
by working professionals from different industries.
XL Your Mind Gneisenaustraße 27 40477 Düsseldorf Germany Email: [email protected] Phone: +49 211 99 54 89 84 Website: www.xlyourmind.com