#### How Algorithms Work

**Tutorial Objective**

To give a general overview of the whole predictive modelling process and the issues you will need to think about for your predictive modelling projects to be successful.

To give code snippets in R to get you on the road to building your first predictive models.

No deep mathematics will be taught, just the concepts of what is going on.

The tutorial will be a small group to ensure there is plenty of interaction – we expect lots of questions.

**Will I need a laptop?**

We will demo most things so you can concentrate on the concepts rather than worrying about what code to write or how to use particular software.

There will be some exercises that you can work through with us if you like, so please bring a laptop along if you want to try them.

The software needed to join in will be:

- R
- R Studio
- Microsoft SQL Server Express – if you want to practice getting data into a database and connecting other software to it using ODBC

Also, please have the following R packages installed by running the code below in R Studio (your homework if you have not used R before).

install.packages(‘nnet’)

install.packages(‘randomForest’)

install.packages(‘gbm’)

install.packages(‘xgboost’)

install.packages(‘snowfall’)

install.packages(‘rlecuyer’)

install.packages(‘RODBC’)

install.packages(‘caTools’)

**What experience level is this course aimed at?**

The course would be suitable for those who have recently started their model building career.

**Some Topics that maybe covered:**

*The pre-modelling*

- Receiving Data
- Data sanity checking
- Problem definition
- Terminology
- Best ways to learn

*The algorithm family tree*

- Linear regression
- Logistic regression
- Multivariate Adaptive Regression Splines
- Neural Networks
- Decision trees

*The other things*

- Model accuracy metrics
- Feature Engineering
- Backpropagation
- Ensembling
- Boosting
- Bagging
- Variable importance