jeenalee / pmml_collection

Collection of pmml files featuring Iris data for reference.

PMML Collection

This is a collection of predictive models based on Iris data set, represented in PMML. The models in R and sklearn were represented in PMML via [R2PMML] (https://github.com/jpmml/r2pmml) and [sklearn2pmml] (https://github.com/jpmml/sklearn2pmml).

Linear Regression

x = Sepal.Width, Petal.Length, Petal.Width

y = Sepal.Length

R: lm {stats}

sklearn: sklearn.linear_model.LinearRegression()

Logistic Regression

x = Sepal.Length, Sepal.Width

y = Species (only 2 states; 'setosa' and 'versicolor')

R: glm {stats} with family='binomial'

sklearn: sklearn.linear_model.LogisticRegression()

R creates a model for each categorical state. For example, if there were two possible states in the data (A or B), R would create two models: one that would predict the likelihood of A, and another for likelihood of B. In sklearn, the decision function calculates the likelihood of state A (probability is closer to 1) or state B (probability is closer to 0).

Random Forest

Also includes models trained on only 10 samples to make simpler models.

x = Sepal.Length, Sepal.Width

y = Species

R: randomForest {randomForest} with ntree=2

sklearn: sklearn.ensemble.RandomForestClassifer(n_estimators = 2)

R creates the trees in breadth-first manner, i.e. the node ids will be continuous in a breadth-first traversal. Meanwhile, sklearn outputs the trees in depth-first manner.