Weekly #105: AutoViz and AutoViML: Automated Visualization and Machine Learning

I will describe what is available in terms of Open Source and Proprietary tools for automating Data Science tasks and introduce 2 new tools: one to visualize any sized data set with one click, another: to try multiple ML models and techniques with a single call. I will provide the Github Repos for both for free in the talk.

[AICamp]

1/3:
10:21:50 From AICamp US : thank you for joining us today, can I do a quick poll which cities/countries you all join from? thanks
10:22:02 Seattle
10:22:03 London
10:22:06 London, UK
10:22:07 Berlin
10:22:07 Portland, OR
10:22:07 Austin, TX
10:22:11 Portland
10:22:19 Nairobi - Kenya
10:22:19 Dallas, TX - US
10:22:20 Paris
10:22:25 istanbul/turkey
10:22:26 Madrid/Spain
10:22:26 Amsterdam
10:22:27 Scottsdale, Arizona US
10:22:31 Paris in France
10:22:31 Toronto, Canada
10:22:44 UK
10:22:46 Fremont, CA
10:23:04 SLC, Utah, USA
10:23:06 Bogota/Colombia
10:23:15 Toronto, Canada
10:23:17 Atlanta/US
10:23:33 Cambridge, UK:-)
10:23:50 Netherlands
10:24:04 Tunis/Tunisia
10:24:08 Boston
10:24:23 From AICamp US : thank you all join from all over the world :-).
10:24:50 From AICamp US : for questions to speaker, you can type here prefix with Q:
10:25:30 Q: How are the most important features selected ? PCA ?
10:25:48 São Paulo/Brazil
10:26:08 London\UK
10:36:39 From Ram Seshadri To All Panelists : datapath = = 'boston.csv'
sep = '\t'
target = 'MEDV'
10:37:13 Could you please paste here the GitHub link?

[lancen.]

Thanks very much for this. It would be much easier to demo this out of Google Colab.

[AICamp]

2/3:


Data Science Bowl 2019
Boston Housing
Wisconsin Breast Cancer
10:38:29 From Ram Seshadri To All Panelists : !pip install autoviz
10:38:56 From Ram Seshadri To All Panelists : from autoviz.AutoViz_Class import AutoViz_Class
AV = AutoViz_Class()
10:42:53 From Ram Seshadri To All Panelists : dft = AV.AutoViz(datapath+filename, sep, target, "",
header=0, verbose=1, lowess=False, chart_format='svg')
10:43:28 From Ram Seshadri To All Panelists : !pip install autoviml
10:43:51 From Ram Seshadri To All Panelists : from autoviml.Auto_ViML import Auto_ViML
10:44:59 From Ram Seshadri To All Panelists : m, feats, trainm, testm = Auto_ViML(train, target, test, sample_submission,
scoring_parameter=scoring_parameter,
hyper_param='GS', feature_reduction=True,
Boosting_Flag=False, Binning_Flag=False,
Add_Poly=0, Stacking_Flag=False,
Imbalanced_Flag=False,
verbose=1)
10:45:49 Can the output of auto_viml be use in production straight away, or would we need to tweak anything? (As in can I just copy paste)
10:47:32 How do these tools do with very large datasets?
10:49:05 From Ram Seshadri To All Panelists : and sign in with your Kaggle ID
10:50:36 From Ram Seshadri To All Panelists : train =
10:51:02 From Ram Seshadri To All Panelists : Next install autoviml in Kaggle Notebooks
!pip install autoviml
10:51:29 From Ram Seshadri To All Panelists : from autoviml.Auto_ViML import Auto_ViML
10:51:53 From Ram Seshadri To All Panelists : Paste this in an empty cell in Kaggle NB

[appuvidhu]

Thanks very much on this detailed video about both the libraries. Could you please give me the code for saving all images from Autoviz library to my local drive as PNG format ? Thanks.

[AICamp]

3/3:
10:51:54 From Ram Seshadri To All Panelists : m, feats, trainm, testm = Auto_ViML(train, target, test,
sample_submission='',
scoring_parameter='', KMeans_Featurizer=False,
hyper_param='GS', feature_reduction=True,
Boosting_Flag=None, Binning_Flag=False,
Add_Poly=0, Stacking_Flag=False, Imbalanced_Flag=False,
verbose=0)
10:52:31 From Ram Seshadri To All Panelists : m, feats, trainm, testm = Auto_ViML(train, target='accuracy_group', test='',
sample_submission='',
scoring_parameter='', KMeans_Featurizer=False,
hyper_param='GS', feature_reduction=True,
Boosting_Flag=None, Binning_Flag=False,
Add_Poly=0, Stacking_Flag=False, Imbalanced_Flag=False,
verbose=0)
10:54:26 From Ram Seshadri To All Panelists : Run a CatBoost Model by doing this:
m, feats, trainm, testm = Auto_ViML(train, target='accuracy_group', test='',
sample_submission='',
scoring_parameter='', KMeans_Featurizer=False,
hyper_param='GS', feature_reduction=True,
Boosting_Flag="CatBoost", Binning_Flag=False,
Add_Poly=0, Stacking_Flag=False, Imbalanced_Flag=False,
verbose=0)
10:55:49 Are about visucalizations built into this Kaggle demo
10:56:38 So this automatically tunes the parameters within the model? Does it show what the final chosen parameters are somewhere?
10:57:06 Q: So this automatically tunes the parameters within the model? Does it show what the final chosen parameters are somewhere?
10:57:51 Can we play with the hyperparameters?
10:58:39 If the dataframe has many null or NaN, will it remove the the record or fill it up with, say mean value of the columns?
10:59:25 What kind of models are available to choose from?
10:59:53 4 kinds of models: Linear, Forest, XGB, CatBoost
11:00:13 Any plans on unsupervised learning?
11:00:24 Q: what models for time series are you thinking of ? Thanks
11:01:06 Thank you very much
11:01:08 Hi Ram, Thank You for the presentation..
11:01:08 Thanks! :)
11:01:08 Thank you.
11:01:08 Thano you
11:01:14 Thank you
11:01:15 thank you, nice tool
11:01:17 Thanks Very much
11:01:20 thank you.
11:01:30 Thanks, good tool
11:01:30 thank you guys, very good presentation
11:01:36 Thank you 👍
11:01:46 amazing. Thank you so, much
11:01:47 Thank you!
11:01:48 Thank you
11:01:49 Possibly insert links in the youtube recording.
11:01:58 Tx!

[emmanuelkolawole]

File encoding decoder utf-8 does not work for this file
File encoding decoder iso-8859-11 does not work for this file
File encoding decoder cpl252 does not work for this file
File encoding decoder latin1 does not work for this file
None of the decoders work...
Not able to read or load file. Please check your inputs and try again...