Basic H2O for Python with Eric Eckstrand
4 likes1,847 views
This document provides an overview of using the H2O machine learning library in Python. It covers getting started with H2O, its basic architecture where Python connects to an H2O JVM, loading sample iris data into the H2O JVM, exploring and manipulating the data within H2O frames, building a k-means clustering model on the iris data within H2O and Scikit-Learn, and saving/loading H2O models.
![Model-Building
1. H2O K-means
a. h2o_model = h2o.kmeans(x=iris_H2OFrame[:,0:4],
k=3)
b. h2o_model.centers()
2. Scikit Learn
a. from sklearn.cluster import KMeans
b. sk_model = KMeans(n_clusters=3)
c. sk_model.fit(iris_DataFrame.iloc[:,0:4])
d. sk_model.cluster_centers_](https://image.slidesharecdn.com/basich2oforpython-150717192048-lva1-app6892/85/Basic-H2O-for-Python-with-Eric-Eckstrand-13-320.jpg)
Basic H2O for Python with Eric Eckstrand
- 1. Basic H2O for Python Eric C. Eckstrand
- 2. Agenda 1. Getting H2O & Documentation 2. Basic Architecture 3. Loading Data 4. Data Exploration & Munging 5. Model Building 6. Model Saving & Loading
- 3. Getting H2O & Docs 1. pip install h2o 2. http://h2o.ai/download/ a. Bleeding Edge (link) b. Install in Python (tab) c. pip install http://h2o-release.s3.amazonaws. com/h2o/master/3066/Python/h2o-3.1.0.3066-py2.py3-none-any.whl 3. build h2o (https://github.com/h2oai/h2o-3#4-building-h2o-3) a. pip install h2o-py/dist/h2o-3.1.0.99999-py2.py3-none-any.whl 4. http://docs.h2o.ai/ -> H2O 3.0 -> Python Users (link) -> Python docs (link)
- 4. Basic Architecture local machine Python >>> import h2o
- 5. Basic Architecture local machine Python >>> import h2o >>> h2o.init() H2O JVM ip=localhost, port=54321
- 6. Basic Architecture local machine Python >>> import h2o >>> h2o.init(ip=”172.16.2.181”, port=54321) H2O JVM ip=172.16.2.181, port=54321 remote machine
- 7. Basic Architecture local machine Python >>> import h2o >>> h2o.init(ip=”172.16.2.181”, port=54321) H2O JVM H2O JVM H2O JVM H2O JVM H2O JVM
- 8. Load Data into H2O JVM 1. Iris dataset a. 150 rows x 5 columns b. Sepal Width, Sepal Length, Petal Width, Petal Length, and Species (Verginica, Setosa, Versicolor) 2. Methods a. h2o.upload_file b. h2o.import_frame c. h2o.H2OFrame
- 9. Load Data into H2O JVM my laptop: /Users/ece/0xdata/h2o-dev/smalldata/iris/iris.csv Python >>> import h2o >>> h2o.init() >>> iris_H2OFrame = h2o.upload_file (“/Users/ece/0xdata/h2o-dev/smalldata/iris/iris.csv”) >>> iris_H2OFrame = h2o.import_frame (“/Users/ece/0xdata/h2o-dev/smalldata/iris/iris.csv”) H2O JVM ip=localhost, port=54321
- 10. Load Data into H2O JVM my laptop: /Users/ece/0xdata/h2o-dev/smalldata/iris/iris.csv Python >>> import h2o >>> h2o.init(ip=”172.16.2.181”, port=54321) >>> >>> iris_H2OFrame = h2o.upload_file (“/Users/ece/0xdata/h2o-dev/smalldata/iris/iris.csv”) >>> >>> iris_H2OFrame = h2o.import_frame (“/home/eric/iris.csv”) H2O JVM ip=172.16.2.181, port=54321 server room: /home/eric/iris.csv
- 11. Exploration & Munging my laptop: /Users/ece/0xdata/h2o-dev/smalldata/iris/iris.csv Python >>> import h2o >>> h2o.init() >>> iris_H2OFrame = h2o.upload_file (“/Users/ece/0xdata/h2o-dev/smalldata/iris/iris.csv”) H2O JVM ip=localhost, port=54321Frame 150 x 5
- 12. Exploration & Munging 1. show, dim, nrow, ncol, head, tail, col_names, setNames 2. indexing 3. summary statistics a. mean, median, min, max, sd 4. categorical columns a. levels 5. cut, group_by 6. ndarray <-> DataFrame <-> H2OFrame
- 13. Model-Building 1. H2O K-means a. h2o_model = h2o.kmeans(x=iris_H2OFrame[:,0:4], k=3) b. h2o_model.centers() 2. Scikit Learn a. from sklearn.cluster import KMeans b. sk_model = KMeans(n_clusters=3) c. sk_model.fit(iris_DataFrame.iloc[:,0:4]) d. sk_model.cluster_centers_
- 14. Model Saving & Loading 1. path = h2o.save_model(h2o_model," /Users/ece/") 2. saved_model = h2o.load_model(str(path)) 3. saved_model.centers()
- 15. Questions?