Get Supervised machine learning: optimization framework and applications with SAS and R First Edition Kolosova PDF ebook with Full Chapters Now
Get Supervised machine learning: optimization framework and applications with SAS and R First Edition Kolosova PDF ebook with Full Chapters Now
com
https://textbookfull.com/product/supervised-machine-
learning-optimization-framework-and-applications-with-sas-
and-r-first-edition-kolosova/
OR CLICK BUTTON
DOWNLOAD NOW
https://textbookfull.com/product/machine-learning-foundations-
supervised-unsupervised-and-advanced-learning-taeho-jo/
textboxfull.com
https://textbookfull.com/product/fundamentals-of-optimization-theory-
with-applications-to-machine-learning-gallier-j/
textboxfull.com
https://textbookfull.com/product/fundamentals-of-optimization-theory-
with-applications-to-machine-learning-gallier-j-2/
textboxfull.com
https://textbookfull.com/product/introduction-to-machine-learning-
with-r-rigorous-mathematical-analysis-first-edition-burger/
textboxfull.com
Applied Analytics through Case Studies Using SAS and R:
Implementing Predictive Models and Machine Learning
Techniques Deepti Gupta
https://textbookfull.com/product/applied-analytics-through-case-
studies-using-sas-and-r-implementing-predictive-models-and-machine-
learning-techniques-deepti-gupta/
textboxfull.com
https://textbookfull.com/product/supervised-learning-with-quantum-
computers-maria-schuld/
textboxfull.com
https://textbookfull.com/product/supervised-machine-learning-in-wind-
forecasting-and-ramp-event-prediction-1st-edition-harsh-s-dhiman/
textboxfull.com
https://textbookfull.com/product/blockchain-big-data-and-machine-
learning-trends-and-applications-first-edition-kumar/
textboxfull.com
https://textbookfull.com/product/teaching-learning-based-optimization-
algorithm-and-its-engineering-applications-1st-edition-r-venkata-rao/
textboxfull.com
Supervised Machine
Learning
Supervised Machine
Learning
Optimization Framework and
Applications with SAS and R
This book contains information obtained from authentic and highly regarded sources. While all
reasonable efforts have been made to publish reliable data and information, neither the author[s]
nor the publisher can accept any legal responsibility or liability for any errors or omissions that may
be made. The publishers wish to make clear that any views or opinions expressed in this book by
individual editors, authors or contributors are personal to them and do not necessarily reflect the
views/opinions of the publishers. The information or guidance contained in this book is intended
for use by medical, scientific or health-care professionals and is provided strictly as a supplement to
the medical or other professional’s own judgement, their knowledge of the patient’s medical history,
relevant manufacturer’s instructions and the appropriate best practice guidelines. Because of the
rapid advances in medical science, any information or advice on dosages, procedures or diagnoses
should be independently verified. The reader is strongly urged to consult the relevant national drug
formulary and the drug companies’ and device or material manufacturers’ printed instructions, and
their websites, before administering or utilizing any of the drugs, devices or materials mentioned in
this book. This book does not indicate whether a particular treatment is appropriate or suitable for
a particular individual. Ultimately it is the sole responsibility of the medical professional to make
his or her own professional judgements, so as to advise and treat patients appropriately. The authors
and publishers have also attempted to trace the copyright holders of all material reproduced in this
publication and apologize to copyright holders if permission to publish in this form has not been
obtained. If any copyright material has not been acknowledged please write and let us know so we
may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.
com or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA
01923, 978-750-8400. For works that are not available on CCC please contact mpkbookspermissions@
tandf.co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Typeset in Palatino
by codeMantra
Dedication
Part I
1. Introduction to the AI Framework .............................................................. 3
1.1 Components of the AI Framework and Their Interaction...............3
1.2 AI Framework in Detail ....................................................................... 5
1.2.1 Creating Training and Test Datasets..................................... 5
1.2.2 Design of Experiments for a Classifier ................................. 6
1.2.3 Firth Logistic Regression ........................................................6
1.2.4 Data
Contamination ................................................................6
1.2.5 Best Classifiers .......................................................................... 7
1.3 SAS Procedures for the AI Framework Components ...................... 7
1.4 R Libraries for the AI Framework Components ............................... 7
References .........................................................................................................8
vii
viii Contents
Part II
Part III
10.1 Introduction
....................................................................................... 115
10.2 Review of Underwriters’ Performance .......................................... 116
10.2.1 Metrics of Underwriters’ Performance ............................. 116
10.2.1.1 Hit Ratio ................................................................. 116
Contents xi
10.2.1.2 Conversion
Rate .................................................... 116
10.2.1.3 Dynamic Conversion Rate .................................. 117
10.2.1.4 Time-to-Deal ......................................................... 118
10.2.2 Analysis of Underwriters’ Performance ........................... 119
10.2.2.1 Data Description................................................... 119
10.2.2.2 Application Flow .................................................. 119
10.2.2.3 Dynamic Conversion Rate per Underwriter .... 121
10.2.2.4 Time-to-Deal per Underwriter ........................... 122
10.3 Traditional Approach to Knowledge Delivery ............................. 123
10.4 Anatomy of Artificial Intelligence Solution .................................. 124
10.4.1 Data Structure ...................................................................... 124
10.4.2 Classification Approach ...................................................... 125
10.4.3 Bias–Variance Trade-Off and SVM Hyperparameters ... 125
10.4.4 Building the Classifier ......................................................... 127
10.4.5 “Contamination” of Training Datasets ............................. 130
10.4.6 Experimental Results .......................................................... 130
10.5 Summary ............................................................................................ 132
References ..................................................................................................... 132
11.1
Introduction ....................................................................................... 135
11.2 Data ..................................................................................................... 136
11.3 The Cox Model for Claims Event Analysis ................................... 136
11.4 Application of the Cox Model for Claims Analysis ..................... 138
11.4.1 Data Transformation ........................................................... 139
11.4.2 Cox Model Assumption Validation ................................... 141
11.4.3 Bayesian Machine Learning Approach ............................ 144
11.4.4 Deployment with SAS ......................................................... 144
11.4.5 Interpretation of Results ..................................................... 146
11.5 Summary ............................................................................................ 152
References ..................................................................................................... 153
Writing this book was a real adventure. It summarizes our experience over
many years of solving problems related to applying machine learning meth-
ods to real-life data. During these years, we met many people who, by sharing
their experience, asking questions, and challenging our solutions, helped to
crystalize this book’s ideas. Among them, Samuel would like to thank Maurice
“Hank” Greenberg for being a unique inspiration and Charles Dangelo for
his expert advice in the insurance domain. Tanya is grateful for the knowl-
edge of the design of experiments shared by Prof. David M. Steinberg from Tel
Aviv University, Israel, and Prof. Carl Schwarz from Simon Fraser University,
Canada. We want to thank the SAS Institute as our long-term software of
choice, as well as the R Core Team and contributors.
We would also like to thank everyone on the Taylor and Francis/CRC
Press team for their support and assistance in preparing this book. Special
thanks to David Grubbs, the acquisition editor, for his great support. Thanks
to Rebecca, Sofia, Varun, and other editorial staff members for helping this
book come to fruition.
Special thanks to our daughter Efrat for her performances of Chopin
and Debussy’s piano works that provided a calming and inspirational
atmosphere.
xiii
Authors
xv
Introduction: Challenges in the
Application of Machine Learning
Classification Methods
xvii
xviii Introduction
( x1 , y1 ) , …, ( xn , yn )
where each case ( xi , y i ) , i = 1, … , n consists of an object xi (a vector of fea-
tures) and its label y i. The problem is to predict the label y n+1 of a new object
xn+1 . Usually, the goal of classification is to produce a prediction ŷ n+1 that is
likely to coincide with the true label y n+1, and this goal should be comple-
mented with some measure of its reliability. There is a clear tradeoff between
accuracy and reliability: we can improve the former by relaxing the latter
and vice versa.
• individual accuracy metrics are often far from the average accuracy,
• individual accuracy metrics are quite different from one another,
and
• the classification method is very sensitive to which dataset it is
applied.
• bootstrap and
• mixture experiments.
contaminate data and to find the level of contamination that “breaks” the
classifier. This helps to identify robust classifiers that are better positioned
for unseen data.
Challenges of Deployment
The useful life of the machine learning application begins when it is deployed
to production. From this point, application modifications are often required.
These modifications may consume a great deal of effort in recoding and
even redesigning. Generally, the better the machine learning application, the
longer its life. Modifications and maintenance of the application are usually
performed by people other than the original implementers. This problem
is very difficult to solve because the application internals cannot be eas-
ily understood. The code-free framework described in this book addresses
these problems. Generally, the framework outlined in this book addresses
the most fundamental problems of existing machine learning application
development and deployment technologies: the failure to recognize users’
true needs and the inability to develop machine learning solutions quickly
to meet these needs.
Machine learning applications have all the challenges of the traditional
software systems, plus an additional set of machine learning–specific issues –
reproducibility and comparability. These two machine learning–specific
issues can be solved by applying the principles of statistically designed
experiments combined with bootstrap along with testing the robustness
of the machine learning solution through data contamination processes
(flipping or corruption of labels of training dataset). The principles and
applications of designed experiments, bootstrap, and data contamination are
discussed in this book.
Another challenge is the selection of the production classifier. If, for exam-
ple, the criterion of a classifier quality is to maximize sensitivity and specific-
ity, then we may encounter a difficult choice. Let us say, we have classifier A
with a sensitivity of 0.85 and a specificity of 0.88, and classifier B with a sensi-
tivity of 0.87 and a specificity of 0.87 – then which of them should we choose?
In this book, we want to explore a different approach, wherein we do not
need to limit our choice by one classifier. Both classifiers, A and B, have a
very good classification quality. More than that, we can find a tens of clas-
sifiers with very similar sensitivity and specificity. We also know that the
sensitivity and specificity of each classifier will vary depending on the data-
sets they are applied to. So, instead of choosing one classifier, we will use all
classifiers that have high-quality metrics, and the final decision on the classi-
fication result will be based on a voting mechanism using some information
criteria. This approach will immediately improve the classification process.
Introduction xxi
Source Code
The source code for the book can be continuously improved and extended
after the book has been published. The source code is located on GitHub:
https://github.com/smlof/Supervised-Machine-Learning--Optimization-
Framework
Part I
1
Introduction to the AI Framework
3
4 Supervised Machine Learning
FIGURE 1.1
Components of the AI framework.
Introduction to the AI Framework 5
FIGURE 1.2
AI framework process.
6 Supervised Machine Learning
Sillä hetkellä eräs köyhä, huivipää tyttö, joka oli itkien polvistunut
rippituoliin, nousi siitä ylös loistavin silmin. Roma vapisi liikutuksesta.
Vastustamaton halu tunnustaa oli vallannut hänet. Hänen täytyi
tarttua kiinni istuimeen, sillä muuten hän olisi kaatunut. Sitten hän
äkillisen tunnelman valtaamana, jonkinmoisessa huumaustilassa ja
tuskin tietäen mitä teki, heittäytyi samalle paikalle, josta tyttö oli
noussut, ja sanoi sykkivin sydämin ja vapisevalla äänellä pienen
metalliristikon läpi:
»Isä, minä olen suuri syntinen — kuule minua, kuule minua!»
»Tyttäreni, sinä et ole tässä asiassa tehnyt syntiä eikä sinulla ole
mitään kaduttavaa. Se, että tuskat sinua vaivaavat, todistaa, että
sielusi on puhdas ja että elät yhteydessä Jumalan kanssa.
Hermostuminen ja huoli on heikontanut ruumiillista terveyttäsi, ja
siksi on luonnollista, että luulet tehneesi syntiä, vaikka et ole
tehnytkään. Se on suloinen ominaisuus useimmissa naisina, mutta
valitettavasti vain harvoissa miehissä! Synti ei ollut sinun syntisi,
mene siis kotiisi rauhassa, ja Jumala sinua lohduttakoon.»
Sinun D. R.»
Davido Rossi.»
XIII.
Armas, kuinka minä saatan jatkaa? Etkö voi arvata, mitä tahtoisin
sanoa sinulle? Kirjeesi pakottaa minut tunnustamaan. Tulkoon mitä
tahansa, en voi vaieta enää. Etkö arvannut kuka ystäväraukkani
on? Arvelin, että muistaisit edellisen kirjeenvaihtomme, kun sinä olit
rakastavinasi jotakin toista. Sinä et nähtävästi ole ajatellut sitä, ja
se on taas todistus — katkera ja suloinen todistus rakkaudestasi ja
luottamuksestasi minuun. Sinä asetit minut niin korkealle, ettet
ollenkaan epäillytkään, että puhuisin itsestäni. Niin oli kumminkin
laita, ja ystäväraukkani olen minä itse.
Vihdoin tuli rakkaus, ja minä tahdoin pitää siitä kiinni. Voi kuinka
hartaasti tahdoin! Siitä syystä en kertonut sinulle. Se oli
jonkinmoista peliä, se oli huumausta. Kaikkea, mitä tapahtui, pidin
rangaistuksena. Tulipa köyhyys, häpeä, kurjuus, entä sitten! Se
puhdistaisi vain syntistä entisyyttäni ja veisi minut lähemmäksi
sinua. Mutta kun vihdoin hän, joka oli loukannut minua, uhkasi
loukata sinua minun kauttani, jouduin epätoivoon. Sinä et aavista,
mitä aikeita silloin haudoin. Aioin surmatakin itseni saadakseni
kaikki loppumaan. Mutta minä en hennonut särkeä sydäntäsi siten.
Paitsi sitä tuo teko jo olisi ilmaissut sinulle jotain, ja minua hirvitti
ajatus, että sinä minun kuoltuani saisit tietää koko kurjan kohtaloni.
Roma.»
XIV.
Nyt, kun hän asui Rossin ympäristössä, tuntui hänestä kuin hän
huomaisi jotain uutta Rossissa joka hetki. Katolla hyppelevät oravat
toivat mieleen Rossin pienenä poikana, ja lintuset, jotka juuri pesivät
ja siitä syystä laulelivat pitkin päivää, panivat Roman ajattelemaan
sykkivin sydämin heitä molempia. Lahjat, joita toiset naiset olivat
antaneet Rossille, herättivät Romassa melkein kuumemaista
uteliaisuutta. Muutamat olivat Englannista, toiset Amerikasta, ja
monet olivat naisilta, jotka eivät koskaan olleet Davido Rossia
nähneetkään. Ne tekivät Roman onnelliseksi ja ylpeäksi, mutta
myöskin hiukan mustasukkaiseksi.
Trinità dei Montilta Roma oli katsellut kaikkea tuota kuin aitiosta
katsellaan näyttämöä, mutta nyt hän oli itse sen keskellä. Tässä
samassa ilmanalassa Davido Rossi eli. Rossi oli ehkä pakosta
joutunut siihen, mutta hän jäi sinne vapaasta tahdosta. Nuori
arkatuntoinen, hienostunut nainen kärsi siellä tuskia, mutta hän
luulotteli olevansa tyytyväinen.
Kaikkialla oli Rossi ja yhä vain Rossi! Joka ilta, kun Roma meni
levolle köyhässä asunnossaan, hänen viimeinen ajatuksensa oli
rakkauden sanelema rukous yön pimeydessä. Tuo rukous oli hyvin
yksinkertainen ja lapsellinen ja sisälsi sen, että Rossi aina rakastaisi
häntä, olipa hän minkälainen tahansa ja sanoipa maailma mitä
tahansa ja tekivätpä pahat ihmiset mitä tahansa.
»Aioin juuri kirjoittaa teille, hyvä neiti», sanoi tuo suuri mies
vaipuen tuoliinsa. »Ikävä kyllä on työni ollut turha. Ei maksa vaivaa
jatkaa. Mies on tunnustanut.»
»Liittolaisensa?»
Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.
textbookfull.com