Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Exploratory Hour #41 データを集計したい
Search
Takato Shiroto
December 03, 2019
Science
0
91
Exploratory Hour #41 データを集計したい
2019/12/3(水)に行ったExploratory Hourのスライドです。
このスライドでは、集計 (Summarize)の基本的な使い方とよく使う関数について紹介しています。
Takato Shiroto
December 03, 2019
Tweet
Share
More Decks by Takato Shiroto
See All by Takato Shiroto
Exploratory v6.7の紹介
takatoshiroto
0
900
Exploratory v6.6の紹介
takatoshiroto
0
1.6k
Exploratory v6.5の紹介
takatoshiroto
0
5.1k
コンバージョン率と信頼区間の推移を可視化する方法
takatoshiroto
1
310
Exploratory Hour #104 - 別の列の値をもとに、カテゴリー列の値の順序を指定したい
takatoshiroto
0
190
Exploratory Hour #105 - 元のデータ順をもとに、カテゴリー列の値の順序を指定したい
takatoshiroto
1
240
Exploratory Hour #102 - complete関数を使って2つの時間の間の値を生成したい
takatoshiroto
0
110
Exploratory Hour #103 - 仕事の開始・終了時間データから、どの時間に何人働いているか知りたい
takatoshiroto
0
100
Exploratory v6.4の紹介
takatoshiroto
0
5.8k
Other Decks in Science
See All in Science
ほたるのひかり/RayTracingCamp10
kugimasa
1
540
Pericarditis Comic
camkdraws
0
1.5k
観察研究における因果推論
nearme_tech
PRO
1
160
Analysis-Ready Cloud-Optimized Data for your community and the entire world with Pangeo-Forge
jbusecke
0
130
いまAI組織が求める企画開発エンジニアとは?
roadroller
2
1.4k
統計的因果探索: 背景知識とデータにより因果仮説を探索する
sshimizu2006
3
690
The thin line between reconstruction, classification, and hallucination in brain decoding
ykamit
1
1.2k
Improving Search @scale with efficient query experimentation @BerlinBuzzwords 2024
searchhub
0
270
How were Quaternion discovered
kinakomoti321
2
1.2k
構造設計のための3D生成AI-最新の取り組みと今後の展開-
kojinishiguchi
0
900
創薬における機械学習技術について
kanojikajino
16
5k
地表面抽出の方法であるSMRFについて紹介
kentaitakura
1
350
Featured
See All Featured
BBQ
matthewcrist
87
9.5k
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
30
4.6k
What’s in a name? Adding method to the madness
productmarketing
PRO
22
3.3k
Distributed Sagas: A Protocol for Coordinating Microservices
caitiem20
330
21k
[Rails World 2023 - Day 1 Closing Keynote] - The Magic of Rails
eileencodes
33
2.1k
10 Git Anti Patterns You Should be Aware of
lemiorhan
PRO
656
59k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
32
2.2k
Imperfection Machines: The Place of Print at Facebook
scottboms
267
13k
Being A Developer After 40
akosma
89
590k
Making Projects Easy
brettharned
116
6k
The Psychology of Web Performance [Beyond Tellerrand 2023]
tammyeverts
46
2.4k
How to Think Like a Performance Engineer
csswizardry
22
1.4k
Transcript
EXPLORATORY 1
2 εϐʔΧʔ നށ ܟొ Customer Succes EXPLORATORY ུྺ େֶࡏֶதʹϑʔυϩεΛݮΒͨ͢ΊʹɺֶੜஂମΛ্ཱͪ͛දΛ ΊΔɻͦͷޙɺϏδωεΛΔͨΊʹԽֶϝʔΧʔͷσϡϙϯͱ
ϑʔυςοΫܥελʔτΞοϓͰӦۀͱϚʔέςΟϯάΛܦݧɻΞϓ ϦͷͷͨΊʹσʔλαΠΤϯε͕ඞཁͩͱײ͡ɺΞϓϦʹಛԽ ͨ͠ϢʔβʔͷߦಈੳπʔϧΛ։ൃ͢ΔاۀʹͯɺΞϓϦۀքͷ KPIੳͳͲΛ୲͢ΔɻݱࡏExploratory, Inc. ͰΧελϚʔαΫη εΛ୲͢ΔΒɺσʔλͷՄࢹԽͱ୳ࡧతσʔλੳΛઐͱͯ͠ σʔλαΠΤϯεͷීٴʹऔΓΉɻ @ShirotoTakato
Exploratory Hour
ࠓिͷ࣭ σʔλΛूܭ͍ͨ͠
Customer Name Sales Mike 34 Mike 26 Mike 36 5
ܭࢉΛ࡞ͷ߹1ߦ1ߦʹܭࢉ݁ՌΛฦ͢ ྫɿܭࢉΛ࡞ʢMutateʣ Customer Name Counts Sales_avg Mike 3 32 Mike 3 32 Mike 3 32
Customer Name counts Sales_avg Mike 3 32 Customer Name Sales
Mike 34 Mike 26 Mike 36 6 ूܭͷ߹ɺάϧʔϓ͝ͱʹ·ͱΊͯ1ߦʹܭࢉ݁ՌΛฦ͢ ྫɿूܭʢSummarizeʣ
7 ूܭʢSummarizeʣ
ؔ ఠཁ sum άϧʔϓͷதͷͷ߹ܭ n άϧʔϓͷதͷߦ n_distinct άϧʔϓͷதͷҰҙͷͷ mean άϧʔϓͷฏۉ
median άϧʔϓͷதԝ min άϧʔϓͷ࠷খ max άϧʔϓͷ࠷େ first άϧʔϓͷ࠷ॳͷ last άϧʔϓͷ࠷ޙͷ ूܭ ؔ 8
ؔ ఠཁ nth άϧʔϓͷN൪ͷ sd άϧʔϓͷඪ४ภࠩ var άϧʔϓͷࢄ IQR άϧʔϓͷ̐Ґൣғ(75%͔Β25%·Ͱ)
mad άϧʔϓͷฏۉઈରภࠩ na_count άϧʔϓͷܽଛͷ na_percent άϧʔϓͷܽଛͷׂ߹ ूܭ ؔ 9
• جຊతͳूܭؔʢߦͷɺ߹ܭɺฏۉɺҰҙͳͷʣ • ࠷සΛmodeؔͰٻΊΔ • ࠷ॳͷʢ࠷ޙͷʣͱ࠷খʢ࠷େʣͷҧ͍ • ϩδΧϧܕͷूܭؔʢTRUEͷɺTRUEͷׂ߹ʣ ूܭ (Summarize)
ച্σʔλ
CustomerID͝ͱʹ • จ݅ • ߪങͨ͠ͷछྨ • ച্߹ܭ • ࠃ •
ސ٬ͷ։࢝ • ฦ ूܭ͍ͨ͠
ूܭ͍ͨ͠
CustomerID͝ͱʹ • จ݅ • ߪങͨ͠ͷछྨ • ച্߹ܭ • ࠃ •
ސ٬ͷ։࢝ • ฦ ूܭ͍ͨ͠
15 εςοϓͷྻϔομϝχϡʔ͔Βूܭ (Summarize)Λબ͢Δ
16 ूܭͷμΠΞϩά͕දࣔ͞Εͨ
17 ʹߦͷΛબ͢Δ ϓϨϏϡʔը໘ʹબ͞Ε͕ͨදࣔ͞ ΕΔɻαϯϓϧ͞Εͨ5000ߦΛදࣔ͢Δ ͜ͱ͕Ͱ͖Δɻ
18 άϧʔϓԽʹCustomer IDΛબ͢Δ Customer ID͝ͱʹάϧʔϓԽͯ͠ɺ ߦͷΛूܭ͢Δ͜ͱ͕Ͱ͖ͨɻ
CustomerID͝ͱʹ • จ݅ • ߪങͨ͠ͷछྨ • ച্߹ܭ • ࠃ •
ސ٬ͷ։࢝ • ฦ ूܭ͍ͨ͠
20 ͲͷΑ͏ʹߪങͨ͠ͷछྨΛ͑Ε͍͍͔ʁ ސ٬໊ Mike ϖϯ Mike ϊʔτ Mike ص
Tom ϊʔτ Tom ϖϯ Tom ϊʔτ
21 ͲͷΑ͏ʹߪങͨ͠ͷछྨΛ͑Ε͍͍͔ʁ ސ٬໊ Mike ϖϯ Mike ϊʔτ Mike ص
Tom ϊʔτ Tom ϖϯ Tom ϊʔτ
22 ҰҙͳͷΛ༻͢Δͱɺॏෳͨ͠ΛΧϯτ ͤͣʹछྨͷΛٻΊΔ͜ͱ͕Ͱ͖Δ ސ٬໊ Mike ϖϯ Mike ϊʔτ Mike
ص Tom ϊʔτ Tom ϖϯ Tom ϊʔτ 3छྨ 2छྨ
23 ϓϥεϘλϯΛΫϦοΫͯ͠ΛՃ͢Δ
24 ʹSub-CategoryΛબͼɺूܭؔʹҰҙͳͷ (unique)Λબ͢Δ ߪങͨ͠αϒΧςΰϦʔͷछྨΛओ ܭ͢Δ͜ͱ͕Ͱ͖ͨɻ
CustomerID͝ͱʹ • จ݅ • ߪങͨ͠ͷछྨ • ച্߹ܭ • ࠃ •
ސ٬ͷ։࢝ • ฦ ूܭ͍ͨ͠
26 ʹSalesΛબͼɺूܭؔʹ߹ܭΛબ͢Δ ސ٬͝ͱʹച্ͷ߹ܭΛूܭ͢Δ ͜ͱ͕Ͱ͖ͨ
CustomerID͝ͱʹ • จ݅ • ߪങͨ͠ͷछྨ • ച্߹ܭ • ࠃ •
ސ٬ͷ։࢝ • ฦ ूܭ͍ͨ͠
Customer Name Country Mike China Mike Japan Mike Japan Tom
US Tom Japan Tom US Customer Name Country Mike ??? Tom ??? ࠃΧςΰϦܕͷͨΊूܭͰ͖ͳ͍ʁ
Customer Name Country Mike China Mike Japan Mike Japan Tom
US Tom Japan Tom US Customer Name Country_mode Mike Japan Tom US ࠷සʢmodeʣ
30 ʹCountryΛબͼɺूܭؔʹ࠷සΛબ͢Δ ސ٬ͷ࠷จճ͕ଟ͍ࠃ͕Ճ Θͬͨɻ
CustomerID͝ͱʹ • จ݅ • ߪങͨ͠ͷछྨ • ച্߹ܭ • ࠃ •
ސ٬ͷ։࢝ • ฦ ूܭ͍ͨ͠
ސ٬͕࠷ॳʹߪೖͨ͠ΛٻΊ͍ͨ
33 ʹOrder DateΛબͼɺूܭؔʹ࠷ॳͷ(first)Λબ͢Δ
34 ຊʹ࠷ॳͷͰ͍͍ͷ͔ʁ
35 ͜ͷσʔλจͷঢॱʹฒͼସ͑͞Ε͍ͯͳ͍
࠷ॳͷ vs ࠷খ
Customer Name Order Date Mike 2015-12-3 Mike 2014-7-14 Mike 2015-12-24
Customer Name ։࢝ Mike 2015-12-3 ࠷ॳͷ(first)ͷ߹
Customer Name Order Date Mike 2015-12-3 Mike 2014-7-14 Mike 2015-12-24
Customer Name ։࢝ Mike 2014-7-14 ࠷খ(min)ͷ߹
39 ʹOrder DateΛબͼɺूܭؔʹ࠷খ(min)Λબ͢Δ ސ٬͕࠷ॳʹจ͕ͨ͠ूܭ ͞Εͨɻ
CustomerID͝ͱʹ • จ݅ • ߪങͨ͠ͷछྨ • ച্߹ܭ • ࠃ •
ސ٬ͷ։࢝ • ฦ ूܭ͍ͨ͠
41 ฦͷྻʹTRUE͔FALSEΛͱΔϩδΧϧܕͷྻ
ϩδΧϧܕͷूܭؔ
• TRUEͷ(FALSEͷ) • TRUEͷׂ߹(FASLEͷׂ߹) ϩδΧϧܕͷूܭؔ
Customer Name Returned Mike TRUE Mike FALSE Mike TRUE Customer
Name ฦ Mike 2 TRUEͷͷ߹
Customer Name Returned Mike TRUE Mike FALSE Mike TRUE Customer
Name ฦ Mike 0.6666 TRUEͷׂ߹ (%)ͷ߹
46 ʹReturnedΛબͼɺूܭؔʹTRUEͷΛબ͢Δ
ूܭ͢Δ͜ͱ͕Ͱ͖ͨ
Q&A
None
None
None
Contact Email
[email protected]
Twitter Hashtag #ExploratoryHour ͰπΠʔτʂ Twitter @ExploratoryJp Exploratory
Hour https://bit.ly/30odd9q