Titanic exercise
With bigML you can simply apply and evaluate a selection of machine learning algorithms in a standardized framework. For this you do not need any coding experience. Create an account on www.bigml.com
When you enter BigML for the first time you see a blank canvas.:
Please login or if logged in Choose Dashboard
This is the blank canvas.. from here you can start upload data, create datasets, manipulates data, create and evaluate models.
When you want to organize your projects, hover over the 3 dots at the right side of the screen:
Here you can make new projects or alter existing ones.
The normal procedure is to start creating a project and then upload the dataset inside the project to start..
Create new project and then name it “Titanic”
Now the project shows up in your project list…
Click on the name titanic to open the project and start the work on the titanic case
The first step is to get data in. That process is called Creating data sources. On the right you see icons that show you all the possibilities to get data in, the one opens up here are the direct way, other way’s are connecting to databases or elastic search etc.
We create a Data Source by Create Source From URL and the url = s3://grioml/Titanic.csv
Here I created a source file for this exercise, just a plain normal csv file, could also be an excellent file.. for example..
Then the following shows the process of creating models and evaluating them.
Titanic data
Let’s take the titanic data that we used before and fit the following four models on a training version (80% of cases) of that data set.
Grab the data set
We can use the following code block to directly load the data in our workspace:
Create a project:
Data source: s3://grioml/Titanic.csv
Prepare the data / create a dataset
Validation set
Let’s split the titanic data into a training and validation set(80/20). Before we do so, we fix the random number generator seed in order to allow for reproduction of our results. Any seed value will do. My favorite seed is 123.
Modeling
We now go through the four models where we predict Survived from the other features in titanic - with the exception of Name, naturally. If we would use Name, we would fit a zero-residual model: i.e. a model for every row seperately.
We first need to set the target for ease of the next steps
Logistic regression model
Let’s fit the logistic regression model
First take out the name as a preferred field, BigML does encode all text and categorical fields so they will automatically be taken into account when modeling.
Try out modeling with name in the set and you will see it tries to fit towards names. Keeping it out gives the following result:
And Evaluating the model gives these results:
The simple model gives this result AUC 0.8194 :
Random forest
Let’s train the random forest:
And look at evaluations
OptiML with cross validations
Now we see that there is a potential in this dataset, we selected features and created better models ourselves lets see what an optimized version could do:
And get yourself a cup of coffee and read a good book on AI/ML statistics, while BigML keeps experimenting with models and hyperparameters to find the optimal model.
Doing cross-validations / trees / regressions / neural nets and forest boosted aranomized etc.
