【GitHub】Practical Machine Learning with H2O

相似文件 换一批

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

h2o
This is the code and data repository for the "Practical Machine Learning with H2O" book, published by O'Reilly.
As well as direct from O'Reilly, it is also available in places like Amazon US and Amazon UK  ( Print ISBN: 978-1-4919-6460-6;   Ebook ISBN: 978-1-4919-6454-5 )
If you find any errors in the code found here, please file a bug report here. For mistakes or typos in the text of the book, use the book's errata page
If you have written your own code that gets a better result on one of the data sets in the book, and want to share, please submit a pull request by putting your contrib folder. Please make all contributions as public domain, or under the MIT or BSD open source license. Remember to include any additional data sets used (or links to them if they are already open and online somewhere stable).

本帖隐藏的内容

Practical Machine Learning with H2O-Master.zip (42.12 MB)

2017-4-23 04:03:09 上传

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏2 回帖

关键词：Practical Learning practic machine earning available published mistakes written please

Files by book chapter

###chapter 1
iris_deeplearning.py
iris_deeplearning.R
iris_deeplearning.reproducible.py
iris_deeplearning.reproducible.R


###chapter 2
drop_and_assign.py
drop_and_assign.R
load_from_dictionary.py
modify_columns.R


###chapter 3
analyze_football_bet_correlations.R
building_energy_correlations.py
load.building_energy.py
load.building_energy.R
load.building_energy.reproducible.py
load.building_energy.reproducible.R
load.football_2013_2014.py
load.football_2013_2014.R
load.football2.py
load.football2.R
load.football.py
load.football.R
load.mnist.py
load.mnist.R
load.mnist.reproducible.py
load.mnist.reproducible.R
mnist_manipulation.py


###chapter 4
early_stopping_example.py
early_stopping_example.reproducible.py
misbalanced_classes.example.R
misbalanced_classes.reproducible.R
using_checkpoints.py
using_checkpoints.R


###chapter 5
compare_models.R
random_forest.building_energy.default.py
random_forest.building_energy.default.R
random_forest.building_energy.default.reproducible.py
random_forest.building_energy.grid1.py
random_forest.building_energy.grid1.R
random_forest.building_energy.grid2.R
random_forest.building_energy.lapply_seed.R
random_forest.football.default.py
random_forest.football.default.reproducible.py
random_forest.mnist.default.py
random_forest.mnist.default.R
random_forest.mnist.default.reproducible.py


###chapter 6
gbm.building_energy.default.R
gbm.building_energy.default.reproducible.py
gbm.building_energy.default.reproducible.R


###chapter 9
autoencoder.tfidf.py
autoencoder.tfidf.R
glrm.tfidf.R
kmeans.tfidf.py
kmeans.tfidf.R
prcomp.tfidf.R


###chapter 11
epilogue_comparison.building_energy.R
epilogue_comparison.football.R
epilogue_comparison.mnist.R

1. #Run the load.building_energy.py script, to set `train`
   
2. 
   
3. import pandas as pd
   
4. import numpy as np
   
5. import matplotlib.pyplot as plt
   
6. 
   
7. #Make the correlations and download to a Pandas frame
   
8. res = train[x].cor(train[y]).as_data_frame()
   
9. res.index = x
   
10. 
    
11. #res.plot.barh();plt.show() is enough to draw the bar chart.
    
12. #All the rest of this code is making it pretty, with labels on each bar.
    
13. 
    
14. ax = res.plot.barh(xlim=[-1.1,+1.15])
    
15. for p in ax.patches:
    
16.   if p.get_x() < 0:
    
17.       ax.annotate("-%.2f" % p.get_width(), (p.get_x() -0.05 , p.get_y()), ha="right", va="center", xytext=(5, 10), textcoords='offset points')
    
18.   else:        
    
19.       ax.annotate("%.2f" % p.get_width(), (p.get_x() + p.get_width() +0.03, p.get_y()), va="center", xytext=(5, 10), textcoords='offset points')
    
20. 
    
21. plt.show()

复制代码

1. import h2o
   
2. h2o.init()
   
3. 
   
4. ###############
   
5. # From dictionary
   
6. 
   
7. patients = {
   
8.   'height':[188, 157, 175],
   
9.   'age':[29, 33, 65],
   
10.   'risk':['A', 'B', 'B']
    
11.   }
    
12. df = h2o.H2OFrame(patients)
    
13. 
    
14. ###############
    
15. # From dictionary, with customization
    
16. 
    
17. patients = {
    
18.   'height':[188, 157, 175.1],
    
19.   'age':[29, 33, 65],
    
20.   'risk':['A', 'B', 'B']
    
21.   }
    
22. df = h2o.H2OFrame.from_python(
    
23.   patients,
    
24.   column_types=['enum', None, None],
    
25.   destination_frame="patients"
    
26.   )
    
27. 
    
28. df.types
    
29. df.frame_id
    
30. 
    
31. 
    
32. ###################
    
33. import pandas as pd
    
34. patients = pd.DataFrame({
    
35.   'height':[188, 157, 175.1],
    
36.   'age':[29, 33, 65],
    
37.   'risk':['A', 'B', 'B']
    
38.   }
    
39. df = h2o.H2OFrame(patients)
    
40. df.types
    
41. df.frame_id
    
42. 
    
43. #############
    
44. patients = pd.DataFrame({
    
45.   'height':[188, 157, 175.1],
    
46.   'age':[29, 33, 65],
    
47.   'risk':['A', 'B', 'B']
    
48.   })
    
49. df = h2o.H2OFrame.from_python(
    
50.   patients,
    
51.   column_names=patients.columns.tolist()
    
52.   )
    
53. df.types
    
54. df.frame_id

复制代码

1. import h2o.grid
   
2. 
   
3. g = h2o.grid.H2OGridSearch(
   
4.   h2o.estimators.H2ORandomForestEstimator(
   
5.     nfolds=10
   
6.     ),
   
7.   hyper_params = {
   
8.     "ntrees": [50, 100, 120],
   
9.     "max_depth": [40, 60],
   
10.     "min_rows": [1, 2]
    
11.     }
    
12.   )
    
13. g.train(x, y, train)

复制代码

1. #seed = 999
   
2. #source("load.building_energy.R")
   
3. #source("makeplot.building_energy_results.R")
   
4. 
   
5. seeds <- c(101, 109, 373, 571, 619, 999)
   
6. 
   
7. defaultModels <- lapply(seeds, function(seed){
   
8.   h2o.randomForest(x, y, train, nfolds = 10, seed = seed)
   
9.   })
   
10. 
    
11. tunedModels <- lapply(seeds, function(seed){
    
12.   h2o.randomForest(x, y, train, nfolds = 10, seed = seed,
    
13.     max_depth = 40, ntrees = 200, sample_rate = 0.7,
    
14.     mtries = 4, col_sample_rate_per_tree = 0.9,
    
15.     stopping_metric = "deviance",
    
16.     stopping_tolerance = 0,
    
17.     stopping_rounds = 5,
    
18.     score_tree_interval = 3)
    
19.   })
    
20. 
    
21. 
    
22. def <- sapply(defaultModels, h2o.rmse, xval = TRUE)
    
23. tuned <- sapply(tunedModels, h2o.rmse, xval = TRUE)
    
24. 
    
25. defT <- sapply(defaultModels, h2o.rmse)
    
26. tunedT <- sapply(tunedModels, h2o.rmse)
    
27. 
    
28. #The box plot on RMSE, cross-validation data
    
29. boxplot(c(def,tuned) ~ c(rep(1,6),rep(2,6)) )
    
30. 
    
31. #Compare with RMSE on training data
    
32. boxplot(c(def,tuned,defT,tunedT) ~ c(rep(1,6),rep(2,6),rep(3,6),rep(4,6)) )
    
33. 
    
34. # Here is same example with labels (also showing how to force factor order!)
    
35. f <- factor(c(rep("Default Model",6),rep("Tuned Model",6)), c("Default Model","Tuned Model"))
    
36. boxplot(c(def,tuned) ~ f, ylim=c(1.75, 1.85), ylab="RMSE")
    
37. 
    
38. #Exported as 900x500
    
39. 
    
40. 
    
41. #Repeat but with MAE instead of RMSE
    
42. def2 <- sapply(defaultModels, h2o.mae, xval = TRUE)
    
43. tuned2 <- sapply(tunedModels, h2o.mae, xval = TRUE)
    
44. 
    
45. defT2 <- sapply(defaultModels, h2o.mae)
    
46. tunedT2 <- sapply(tunedModels, h2o.mae)
    
47. 
    
48. boxplot(c(def2,tuned2,defT2,tunedT2) ~ c(rep(5,6),rep(6,6),rep(7,6),rep(8,6)) )
    
49. 
    
50. #And both RMSE and MAE
    
51. boxplot(c(def,tuned,defT,tunedT,def2,tuned2,defT2,tunedT2) ~ c(rep(1,6),rep(2,6),rep(3,6),rep(4,6),rep(5,6),rep(6,6),rep(7,6),rep(8,6)) )
    
52. 
    
53. # Make the results plot
    
54. mBest <- tunedModels[[1]]  #Choose first one, arbitrarily
    
55. p = h2o.predict(mBest, test, seed = seed)
    
56. p
    
57. 
    
58. plotBuildingEnergy(p, test, forPrint = T, "../chapters/images/", "randomforest.tuned")

复制代码

谢谢分享

谢谢分享。

谢谢分享

h2o
This is the code and data repository for the "Practical Machine Learning with H2O" book, published by O'Reilly.
本文来自: 人大经济论坛 winbugs及其他软件专版 版，详细出处参考： https://bbs.pinggu.org/forum.php?mod=viewthread&tid=5552343&page=1&from^^uid=341570