SlideShare uma empresa Scribd logo
1 de 21
Baixar para ler offline
How to correctly estimate
the effect of online
advertisement
Introduction of Double Machine Lerning
About me
● Yusuke Kaneko(@coldstart_p)
● CyberAgent, Inc.
● Studying: Econometrics, Causal Inference
Main Theme
How to estimate correctly treatment effects in
high-dimensional data(such as advertising data)
Original
1. Chernozhukov, Victor, et al. "Double/debiased/neyman machine learning of
treatment effects." American Economic Review 107.5 (2017): 261-65.
2. Belloni, A., Chernozhukov, V., Fernández‐Val, I., & Hansen, C. (2017).
Program evaluation and causal inference with high‐dimensional data.
Econometrica, 85(1), 233-298.
3. Chernozhukov, Victor, et al. "Double/debiased machine learning for treatment
and structural parameters." The Econometrics Journal 21.1 (2018): C1-C68.
Problem
● Advertisement data has a lot of variables. (= High-Dimensional Data)
example: Type of Advertisement , Time , History-data
→ Under such circumstances, it is known that low-dimensional parameters such
as treatment effects can not be estimated well by Machine Learning Methods
(The model and figure are on the slide later)
● Motivation: How to estimate treatment effects correctly ?
→ Use Double Machine Learning(Double ML) !
=( FLEXIBLE and SIMPLE Method)
● Double ML Code are on github(https://github.com/VC2015/DMLonGitHub/)
…(Double ML package does not exist yet. )
Data
● Criteo Uplift Modeling Dataset(EC Site AD Large Dataset for causal
inference)
Recently
Published
Dataset!
Data
● Features are as follows
● Averege Visit Rate and Treatment Ratio
is on the right
Model
● Partially Linear Model
Y: Outcome Variable(Whether user visited = visit)
D: Treatment Variable (Whether or not an ad has been delivered = treatment)
Z: Vector of covariates(such as history data = f0 f1 ...)
θ0: “lift” parameter(Treatment Effect parameter)
● D is expressed by the following equation
D is conditionally exogenous
Bad ML Estimation
● Predict Y using D and Z, then obtain
(Example)
1. Run Random Forest and get
2. and get
3. Repeat 1. and 2. until convergence
Bad ML Estimation
● Prediction performance is Good, but the distribution of looks like
below.
(from paper)
Double ML
1. Predict Y and D using Z by
obtained using ML methods(Lasso , Random Forest etc... )
2. Residualize
3. Regress and get
Double ML
● Prediction performance is Good, and the distribution of looks like
below.
(from paper)
Sample Splitting
● In Double ML, Sample Splitting is one key Ingredient.
→ Convergence rate improves with sample splitting(Please refer to the papers
for the theoretical content)
(from paper)
Double ML in R
● All the code for the simulation is on the github
Simulation
● Since Criteo Dataset is too large , I made overall sampling first (about 2%)
● Introduce Sampling Bias( Since we are using data coming from a randomized
experiment)
Naive ATE Method
● Unbiased ATE(Average Treatment Effect) estimation is possible with this
technique if there is no sampling bias
Simulation
● Average treatment effect without sampling bias on the left
● Average treatment effect with sampling bias on the right
Highly Biased
Simulation
● Use approach: Double ML with Lasso(You can use RandomForest instead of
Lasso)
→ Counter approach: Naive ATE, Single equation Lasso , Usual Lasso , Lasso
with Propensity Score Weighting
Simulation
● Result(with Lasso with Propensity Score Weighting) Lasso with PSW‘s
performance is
very bad , so
remove this
Simulation
● Result(without Lasso with Propensity Score Weighting)
Double ML’s
performance is
very good!
Conclusion
● Double ML is Flexible and Simple Method for causal inference
● Double ML’s performance seems very good with real data

Mais conteúdo relacionado

Mais procurados

ロジスティック回帰の考え方・使い方 - TokyoR #33
ロジスティック回帰の考え方・使い方 - TokyoR #33ロジスティック回帰の考え方・使い方 - TokyoR #33
ロジスティック回帰の考え方・使い方 - TokyoR #33
horihorio
 
ロジスティック回帰分析の入門 -予測モデル構築-
ロジスティック回帰分析の入門 -予測モデル構築-ロジスティック回帰分析の入門 -予測モデル構築-
ロジスティック回帰分析の入門 -予測モデル構築-
Koichiro Gibo
 
多重代入法の書き方 公開用
多重代入法の書き方 公開用 多重代入法の書き方 公開用
多重代入法の書き方 公開用
Koichiro Gibo
 

Mais procurados (20)

Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15
Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15
Misha Bilenko, Principal Researcher, Microsoft at MLconf SEA - 5/01/15
 
Variational Inference
Variational InferenceVariational Inference
Variational Inference
 
星野「調査観察データの統計科学」第1&2章
星野「調査観察データの統計科学」第1&2章星野「調査観察データの統計科学」第1&2章
星野「調査観察データの統計科学」第1&2章
 
DARM勉強会第3回 (missing data analysis)
DARM勉強会第3回 (missing data analysis)DARM勉強会第3回 (missing data analysis)
DARM勉強会第3回 (missing data analysis)
 
2 3.GLMの基礎
2 3.GLMの基礎2 3.GLMの基礎
2 3.GLMの基礎
 
ベルヌーイ分布における超パラメータ推定のための経験ベイズ法
ベルヌーイ分布における超パラメータ推定のための経験ベイズ法ベルヌーイ分布における超パラメータ推定のための経験ベイズ法
ベルヌーイ分布における超パラメータ推定のための経験ベイズ法
 
検定力分析とベイズファクターデザイン分析によるサンプルサイズ設計【※Docswellにも同じものを上げています】
検定力分析とベイズファクターデザイン分析によるサンプルサイズ設計【※Docswellにも同じものを上げています】検定力分析とベイズファクターデザイン分析によるサンプルサイズ設計【※Docswellにも同じものを上げています】
検定力分析とベイズファクターデザイン分析によるサンプルサイズ設計【※Docswellにも同じものを上げています】
 
多目的遺伝的アルゴリズム
多目的遺伝的アルゴリズム多目的遺伝的アルゴリズム
多目的遺伝的アルゴリズム
 
Google Colaboratoryの使い方
Google Colaboratoryの使い方Google Colaboratoryの使い方
Google Colaboratoryの使い方
 
ベイズ統計学の概論的紹介
ベイズ統計学の概論的紹介ベイズ統計学の概論的紹介
ベイズ統計学の概論的紹介
 
R実践 機械学習による異常検知 01
R実践 機械学習による異常検知 01R実践 機械学習による異常検知 01
R実践 機械学習による異常検知 01
 
猫でも分かりたい線形回帰の自由度
猫でも分かりたい線形回帰の自由度猫でも分かりたい線形回帰の自由度
猫でも分かりたい線形回帰の自由度
 
ロジスティック回帰の考え方・使い方 - TokyoR #33
ロジスティック回帰の考え方・使い方 - TokyoR #33ロジスティック回帰の考え方・使い方 - TokyoR #33
ロジスティック回帰の考え方・使い方 - TokyoR #33
 
ロジスティック回帰分析の入門 -予測モデル構築-
ロジスティック回帰分析の入門 -予測モデル構築-ロジスティック回帰分析の入門 -予測モデル構築-
ロジスティック回帰分析の入門 -予測モデル構築-
 
22 Machine Learning Feature Selection
22 Machine Learning Feature Selection22 Machine Learning Feature Selection
22 Machine Learning Feature Selection
 
Stanの便利な事後処理関数
Stanの便利な事後処理関数Stanの便利な事後処理関数
Stanの便利な事後処理関数
 
東大大学院 電子情報学特論講義資料「深層学習概論と理論解析の課題」大野健太
東大大学院 電子情報学特論講義資料「深層学習概論と理論解析の課題」大野健太東大大学院 電子情報学特論講義資料「深層学習概論と理論解析の課題」大野健太
東大大学院 電子情報学特論講義資料「深層学習概論と理論解析の課題」大野健太
 
ジャーナルが論文をリジェクトする最も一般的な理由
ジャーナルが論文をリジェクトする最も一般的な理由ジャーナルが論文をリジェクトする最も一般的な理由
ジャーナルが論文をリジェクトする最も一般的な理由
 
多重代入法の書き方 公開用
多重代入法の書き方 公開用 多重代入法の書き方 公開用
多重代入法の書き方 公開用
 
ベイズ統計モデリングと心理学
ベイズ統計モデリングと心理学ベイズ統計モデリングと心理学
ベイズ統計モデリングと心理学
 

Semelhante a How to correctly estimate the effect of online advertisement(About Double Machine Learning)

Experimental designs and data analysis in the field of Agronomy science by ma...
Experimental designs and data analysis in the field of Agronomy science by ma...Experimental designs and data analysis in the field of Agronomy science by ma...
Experimental designs and data analysis in the field of Agronomy science by ma...
Manoj Sharma
 
Internship project report,Predictive Modelling
Internship project report,Predictive ModellingInternship project report,Predictive Modelling
Internship project report,Predictive Modelling
Amit Kumar
 

Semelhante a How to correctly estimate the effect of online advertisement(About Double Machine Learning) (20)

Faster and cheaper, smart ab experiments - public ver.
Faster and cheaper, smart ab experiments - public ver.Faster and cheaper, smart ab experiments - public ver.
Faster and cheaper, smart ab experiments - public ver.
 
Introductionedited
IntroductioneditedIntroductionedited
Introductionedited
 
Nbe rcausalpredictionv111 lecture2
Nbe rcausalpredictionv111 lecture2Nbe rcausalpredictionv111 lecture2
Nbe rcausalpredictionv111 lecture2
 
PREDICT 422 - Module 1.pptx
PREDICT 422 - Module 1.pptxPREDICT 422 - Module 1.pptx
PREDICT 422 - Module 1.pptx
 
Dissertation
DissertationDissertation
Dissertation
 
IMAGE CLASSIFICATION USING DIFFERENT CLASSICAL APPROACHES
IMAGE CLASSIFICATION USING DIFFERENT CLASSICAL APPROACHESIMAGE CLASSIFICATION USING DIFFERENT CLASSICAL APPROACHES
IMAGE CLASSIFICATION USING DIFFERENT CLASSICAL APPROACHES
 
Adapting neural networks for the estimation of treatment effects
Adapting neural networks for the estimation of treatment effectsAdapting neural networks for the estimation of treatment effects
Adapting neural networks for the estimation of treatment effects
 
AIML_UNIT 2 _PPT_HAND NOTES_MPS.pdf
AIML_UNIT 2 _PPT_HAND NOTES_MPS.pdfAIML_UNIT 2 _PPT_HAND NOTES_MPS.pdf
AIML_UNIT 2 _PPT_HAND NOTES_MPS.pdf
 
Housing price prediction
Housing price predictionHousing price prediction
Housing price prediction
 
Experimental designs and data analysis in the field of Agronomy science by ma...
Experimental designs and data analysis in the field of Agronomy science by ma...Experimental designs and data analysis in the field of Agronomy science by ma...
Experimental designs and data analysis in the field of Agronomy science by ma...
 
Setting up an A/B-testing framework
Setting up an A/B-testing frameworkSetting up an A/B-testing framework
Setting up an A/B-testing framework
 
Internship project report,Predictive Modelling
Internship project report,Predictive ModellingInternship project report,Predictive Modelling
Internship project report,Predictive Modelling
 
housepriceprediction-180915174356.pdf
housepriceprediction-180915174356.pdfhousepriceprediction-180915174356.pdf
housepriceprediction-180915174356.pdf
 
Machine Learning.pptx
Machine Learning.pptxMachine Learning.pptx
Machine Learning.pptx
 
Data Science - Part V - Decision Trees & Random Forests
Data Science - Part V - Decision Trees & Random Forests Data Science - Part V - Decision Trees & Random Forests
Data Science - Part V - Decision Trees & Random Forests
 
Probability distribution Function & Decision Trees in machine learning
Probability distribution Function  & Decision Trees in machine learningProbability distribution Function  & Decision Trees in machine learning
Probability distribution Function & Decision Trees in machine learning
 
Ijaems apr-2016-23 Study of Pruning Techniques to Predict Efficient Business ...
Ijaems apr-2016-23 Study of Pruning Techniques to Predict Efficient Business ...Ijaems apr-2016-23 Study of Pruning Techniques to Predict Efficient Business ...
Ijaems apr-2016-23 Study of Pruning Techniques to Predict Efficient Business ...
 
Classification of Headache Disorder Using Random Forest Algorithm.pptx
Classification of Headache Disorder Using Random Forest Algorithm.pptxClassification of Headache Disorder Using Random Forest Algorithm.pptx
Classification of Headache Disorder Using Random Forest Algorithm.pptx
 
debatrim_report (1)
debatrim_report (1)debatrim_report (1)
debatrim_report (1)
 
How ml can improve purchase conversions
How ml can improve purchase conversionsHow ml can improve purchase conversions
How ml can improve purchase conversions
 

Mais de Yusuke Kaneko

Mais de Yusuke Kaneko (7)

Kdd 2021 読み会(clustering for private interest-based advertising & learning a l...
Kdd 2021 読み会(clustering for private interest-based advertising & learning a l...Kdd 2021 読み会(clustering for private interest-based advertising & learning a l...
Kdd 2021 読み会(clustering for private interest-based advertising & learning a l...
 
DID, Synthetic Control, CausalImpact
DID, Synthetic Control, CausalImpactDID, Synthetic Control, CausalImpact
DID, Synthetic Control, CausalImpact
 
企業の中の経済学
企業の中の経済学企業の中の経済学
企業の中の経済学
 
TokyoR_74_RDD
TokyoR_74_RDDTokyoR_74_RDD
TokyoR_74_RDD
 
LightGBM: a highly efficient gradient boosting decision tree
LightGBM: a highly efficient gradient boosting decision treeLightGBM: a highly efficient gradient boosting decision tree
LightGBM: a highly efficient gradient boosting decision tree
 
Hastie_chapter5
Hastie_chapter5Hastie_chapter5
Hastie_chapter5
 
効果のあるクリエイティブ広告の見つけ方(Contextual Bandit + TS or UCB)
効果のあるクリエイティブ広告の見つけ方(Contextual Bandit + TS or UCB)効果のあるクリエイティブ広告の見つけ方(Contextual Bandit + TS or UCB)
効果のあるクリエイティブ広告の見つけ方(Contextual Bandit + TS or UCB)
 

Último

Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
HyderabadDolls
 
怎样办理圣地亚哥州立大学毕业证(SDSU毕业证书)成绩单学校原版复制
怎样办理圣地亚哥州立大学毕业证(SDSU毕业证书)成绩单学校原版复制怎样办理圣地亚哥州立大学毕业证(SDSU毕业证书)成绩单学校原版复制
怎样办理圣地亚哥州立大学毕业证(SDSU毕业证书)成绩单学校原版复制
vexqp
 
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
nirzagarg
 
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
gajnagarg
 
Gartner's Data Analytics Maturity Model.pptx
Gartner's Data Analytics Maturity Model.pptxGartner's Data Analytics Maturity Model.pptx
Gartner's Data Analytics Maturity Model.pptx
chadhar227
 
Reconciling Conflicting Data Curation Actions: Transparency Through Argument...
Reconciling Conflicting Data Curation Actions:  Transparency Through Argument...Reconciling Conflicting Data Curation Actions:  Transparency Through Argument...
Reconciling Conflicting Data Curation Actions: Transparency Through Argument...
Bertram Ludäscher
 
Sealdah % High Class Call Girls Kolkata - 450+ Call Girl Cash Payment 8005736...
Sealdah % High Class Call Girls Kolkata - 450+ Call Girl Cash Payment 8005736...Sealdah % High Class Call Girls Kolkata - 450+ Call Girl Cash Payment 8005736...
Sealdah % High Class Call Girls Kolkata - 450+ Call Girl Cash Payment 8005736...
HyderabadDolls
 
In Riyadh ((+919101817206)) Cytotec kit @ Abortion Pills Saudi Arabia
In Riyadh ((+919101817206)) Cytotec kit @ Abortion Pills Saudi ArabiaIn Riyadh ((+919101817206)) Cytotec kit @ Abortion Pills Saudi Arabia
In Riyadh ((+919101817206)) Cytotec kit @ Abortion Pills Saudi Arabia
ahmedjiabur940
 
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
nirzagarg
 
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
ZurliaSoop
 
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
gajnagarg
 

Último (20)

Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
Sonagachi * best call girls in Kolkata | ₹,9500 Pay Cash 8005736733 Free Home...
 
Kings of Saudi Arabia, information about them
Kings of Saudi Arabia, information about themKings of Saudi Arabia, information about them
Kings of Saudi Arabia, information about them
 
Vadodara 💋 Call Girl 7737669865 Call Girls in Vadodara Escort service book now
Vadodara 💋 Call Girl 7737669865 Call Girls in Vadodara Escort service book nowVadodara 💋 Call Girl 7737669865 Call Girls in Vadodara Escort service book now
Vadodara 💋 Call Girl 7737669865 Call Girls in Vadodara Escort service book now
 
怎样办理圣地亚哥州立大学毕业证(SDSU毕业证书)成绩单学校原版复制
怎样办理圣地亚哥州立大学毕业证(SDSU毕业证书)成绩单学校原版复制怎样办理圣地亚哥州立大学毕业证(SDSU毕业证书)成绩单学校原版复制
怎样办理圣地亚哥州立大学毕业证(SDSU毕业证书)成绩单学校原版复制
 
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
Top profile Call Girls In Satna [ 7014168258 ] Call Me For Genuine Models We ...
 
Dubai Call Girls Peeing O525547819 Call Girls Dubai
Dubai Call Girls Peeing O525547819 Call Girls DubaiDubai Call Girls Peeing O525547819 Call Girls Dubai
Dubai Call Girls Peeing O525547819 Call Girls Dubai
 
High Profile Call Girls Service in Jalore { 9332606886 } VVIP NISHA Call Girl...
High Profile Call Girls Service in Jalore { 9332606886 } VVIP NISHA Call Girl...High Profile Call Girls Service in Jalore { 9332606886 } VVIP NISHA Call Girl...
High Profile Call Girls Service in Jalore { 9332606886 } VVIP NISHA Call Girl...
 
SAC 25 Final National, Regional & Local Angel Group Investing Insights 2024 0...
SAC 25 Final National, Regional & Local Angel Group Investing Insights 2024 0...SAC 25 Final National, Regional & Local Angel Group Investing Insights 2024 0...
SAC 25 Final National, Regional & Local Angel Group Investing Insights 2024 0...
 
Predicting HDB Resale Prices - Conducting Linear Regression Analysis With Orange
Predicting HDB Resale Prices - Conducting Linear Regression Analysis With OrangePredicting HDB Resale Prices - Conducting Linear Regression Analysis With Orange
Predicting HDB Resale Prices - Conducting Linear Regression Analysis With Orange
 
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
Top profile Call Girls In bhavnagar [ 7014168258 ] Call Me For Genuine Models...
 
Gartner's Data Analytics Maturity Model.pptx
Gartner's Data Analytics Maturity Model.pptxGartner's Data Analytics Maturity Model.pptx
Gartner's Data Analytics Maturity Model.pptx
 
Reconciling Conflicting Data Curation Actions: Transparency Through Argument...
Reconciling Conflicting Data Curation Actions:  Transparency Through Argument...Reconciling Conflicting Data Curation Actions:  Transparency Through Argument...
Reconciling Conflicting Data Curation Actions: Transparency Through Argument...
 
Gulbai Tekra * Cheap Call Girls In Ahmedabad Phone No 8005736733 Elite Escort...
Gulbai Tekra * Cheap Call Girls In Ahmedabad Phone No 8005736733 Elite Escort...Gulbai Tekra * Cheap Call Girls In Ahmedabad Phone No 8005736733 Elite Escort...
Gulbai Tekra * Cheap Call Girls In Ahmedabad Phone No 8005736733 Elite Escort...
 
Sealdah % High Class Call Girls Kolkata - 450+ Call Girl Cash Payment 8005736...
Sealdah % High Class Call Girls Kolkata - 450+ Call Girl Cash Payment 8005736...Sealdah % High Class Call Girls Kolkata - 450+ Call Girl Cash Payment 8005736...
Sealdah % High Class Call Girls Kolkata - 450+ Call Girl Cash Payment 8005736...
 
Charbagh + Female Escorts Service in Lucknow | Starting ₹,5K To @25k with A/C...
Charbagh + Female Escorts Service in Lucknow | Starting ₹,5K To @25k with A/C...Charbagh + Female Escorts Service in Lucknow | Starting ₹,5K To @25k with A/C...
Charbagh + Female Escorts Service in Lucknow | Starting ₹,5K To @25k with A/C...
 
In Riyadh ((+919101817206)) Cytotec kit @ Abortion Pills Saudi Arabia
In Riyadh ((+919101817206)) Cytotec kit @ Abortion Pills Saudi ArabiaIn Riyadh ((+919101817206)) Cytotec kit @ Abortion Pills Saudi Arabia
In Riyadh ((+919101817206)) Cytotec kit @ Abortion Pills Saudi Arabia
 
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
Top profile Call Girls In Bihar Sharif [ 7014168258 ] Call Me For Genuine Mod...
 
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Surabaya ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
 
Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...
Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...
Digital Advertising Lecture for Advanced Digital & Social Media Strategy at U...
 
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
Top profile Call Girls In Vadodara [ 7014168258 ] Call Me For Genuine Models ...
 

How to correctly estimate the effect of online advertisement(About Double Machine Learning)

  • 1. How to correctly estimate the effect of online advertisement Introduction of Double Machine Lerning
  • 2. About me ● Yusuke Kaneko(@coldstart_p) ● CyberAgent, Inc. ● Studying: Econometrics, Causal Inference
  • 3. Main Theme How to estimate correctly treatment effects in high-dimensional data(such as advertising data)
  • 4. Original 1. Chernozhukov, Victor, et al. "Double/debiased/neyman machine learning of treatment effects." American Economic Review 107.5 (2017): 261-65. 2. Belloni, A., Chernozhukov, V., Fernández‐Val, I., & Hansen, C. (2017). Program evaluation and causal inference with high‐dimensional data. Econometrica, 85(1), 233-298. 3. Chernozhukov, Victor, et al. "Double/debiased machine learning for treatment and structural parameters." The Econometrics Journal 21.1 (2018): C1-C68.
  • 5. Problem ● Advertisement data has a lot of variables. (= High-Dimensional Data) example: Type of Advertisement , Time , History-data → Under such circumstances, it is known that low-dimensional parameters such as treatment effects can not be estimated well by Machine Learning Methods (The model and figure are on the slide later) ● Motivation: How to estimate treatment effects correctly ? → Use Double Machine Learning(Double ML) ! =( FLEXIBLE and SIMPLE Method) ● Double ML Code are on github(https://github.com/VC2015/DMLonGitHub/) …(Double ML package does not exist yet. )
  • 6. Data ● Criteo Uplift Modeling Dataset(EC Site AD Large Dataset for causal inference) Recently Published Dataset!
  • 7. Data ● Features are as follows ● Averege Visit Rate and Treatment Ratio is on the right
  • 8. Model ● Partially Linear Model Y: Outcome Variable(Whether user visited = visit) D: Treatment Variable (Whether or not an ad has been delivered = treatment) Z: Vector of covariates(such as history data = f0 f1 ...) θ0: “lift” parameter(Treatment Effect parameter) ● D is expressed by the following equation D is conditionally exogenous
  • 9. Bad ML Estimation ● Predict Y using D and Z, then obtain (Example) 1. Run Random Forest and get 2. and get 3. Repeat 1. and 2. until convergence
  • 10. Bad ML Estimation ● Prediction performance is Good, but the distribution of looks like below. (from paper)
  • 11. Double ML 1. Predict Y and D using Z by obtained using ML methods(Lasso , Random Forest etc... ) 2. Residualize 3. Regress and get
  • 12. Double ML ● Prediction performance is Good, and the distribution of looks like below. (from paper)
  • 13. Sample Splitting ● In Double ML, Sample Splitting is one key Ingredient. → Convergence rate improves with sample splitting(Please refer to the papers for the theoretical content) (from paper)
  • 14. Double ML in R ● All the code for the simulation is on the github
  • 15. Simulation ● Since Criteo Dataset is too large , I made overall sampling first (about 2%) ● Introduce Sampling Bias( Since we are using data coming from a randomized experiment)
  • 16. Naive ATE Method ● Unbiased ATE(Average Treatment Effect) estimation is possible with this technique if there is no sampling bias
  • 17. Simulation ● Average treatment effect without sampling bias on the left ● Average treatment effect with sampling bias on the right Highly Biased
  • 18. Simulation ● Use approach: Double ML with Lasso(You can use RandomForest instead of Lasso) → Counter approach: Naive ATE, Single equation Lasso , Usual Lasso , Lasso with Propensity Score Weighting
  • 19. Simulation ● Result(with Lasso with Propensity Score Weighting) Lasso with PSW‘s performance is very bad , so remove this
  • 20. Simulation ● Result(without Lasso with Propensity Score Weighting) Double ML’s performance is very good!
  • 21. Conclusion ● Double ML is Flexible and Simple Method for causal inference ● Double ML’s performance seems very good with real data