I have read 13 books that connect data science to business results, so I will write a book review.

Motivation for this article

I can't go out in this age, so I'm reading a book anyway. Look at the items below and aim to help those who have the same feelings with this purchase.

--I have learned (learned) the theory of data science and coding with analysis tools. ――Work (business) is required to have "results of using the model" rather than "completion of high-precision model creation" ――You have to give a concrete number ――I want to know how to run a project without a manager or seniors and connect it to business impact. ――Knowing the necessary elements for organizing data science in-house, you can request and persuade the management side (I wish I could become)

For that purpose, I read a book on ** Business + Data Science **.

If you are interested in reading the introduction below, please buy it.

** * There are some parts that are delicately evaluated due to my low absorption capacity. Please check it with your own eyes without making a cormorant. ** **

They are arranged in the order that they find useful for this purpose.

Arbitrary positional relationship summary

1st book: Data scientist training reader Business utilization edition

Author: Multiple people are in charge of each chapter
the publisher:Technical Review Company
Release Date: 2018/10/30

Impressions

--Data analysis Mook book's famous "training reader series" ――You can cry (handkerchief is required because you have too much idea about failure cases) ――The first word of 1-1 is "I don't like data analysis that can't contribute to business anymore." -** Why data science projects are not successful **, ** Reasons to stop at PoC **, ** Creation of data science organization and membership ** He has given various opinions in each chapter. ――I thought that it was an advertisement that many people who were close to consultants wrote "Let's put in a consultant here", but the disadvantage of "self-reliance" is as you said, so read it without worrying so much. Advanced

Content introduction

――A book in which people with various backgrounds and company history are in charge of each chapter --Written by someone who is a manager of a data science team The following problems and how to work are described. --How to proceed with a data science project --Differences between software and data science project management --What to expect from data scientists --Creating a follow-up system and an environment where you can play an active role --Misunderstanding about data science at the time of hiring --Gap in the meaning of "data analysis" between management / managers and data scientists --A troubled scientist who just reads and implements the latest paper --A troubled manager who cannot be sublimated into a project

――If the team is not in place, it is equivalent to being required to be a versatile person (hero type human resource) ――Do not do it alone. Even a versatile person will be overloaded, so I won't let him do it. --Enter another member in the infrastructure app business. Consulting if necessary. ――The data science method is also violent top-down, saying "If you graduate from the mathematics school, you can solve it." ――The above decides hiring / assignment without knowing the degree of specialization for each method

2nd book: A book to acquire real data analysis ability

Author: Shinichi Kawamura
the publisher:Nikkei BP
Release Date: 2016/6/22

Impressions

――There are many frameworks for solving business problems, but I think this is the only book written with concepts related to data analysis. --Contents of other books "It is important to connect to evaluation without stopping by analysis, and emphasis should be placed on problem definition and its review." --Contents of this book "Let's design from goal setting with few rework without omission, analysis execution, and result expression while using the framework" ――Brainstorming using A4 paper and sticky notes, and precautions for analyzing with data are studded. ――Since neither python nor R is used, the feeling of glittering data science is thin. ――Although you can proceed with excel and A4 paper, be careful because it is not a "book for forcibly advancing problems in a project by machine learning" but a "book for what kind of data analysis is reliable and solves problems".

Content introduction

--Reassuring subtitle: The theory is good, what you want is a model of practice --Framework for problem setting and its simulated experience --Points to note when designing analysis projects ――The purpose and result of "Let's analyze because there is data" are opposite --Grasp the features with "visualization" rather than "aggregation" --Problem setting-> Identify factors-> Drop in a schematic diagram-> Collect and check data (outliers / reliability of collection background)-> Eliminate recognition bias

Third book: The strongest data analysis organization

Author: Kaoru Kawamoto
the publisher:Nikkei BP
Release Date: 2017/11/24

Impressions

――The trouble you are having was written in this book --Analysis team just started up --Data literacy has not penetrated the entire company --There is no company-wide promotion system --A book that can be a role model for those who belong to the above analysis team (especially manager) ―― “The method is not the analysis that precedes it, but the method that is necessary for the analysis that is useful as a company is selected by back calculation. However, it is at least necessary for an analyst to try a new analysis method and learn many methods.”

Content introduction

――It describes a steady bottom-up process from a small research team to growing as a team that reforms operations through data analysis over 18 years. ――The ** "bonus" ** feeling of the data analysis organization at the time of launch --Management to dispel the anxiety of team members ――In-house data analyst needed "** business consulting ability " --Back office type analytical knowledge - Front office type site visit / negotiation / problem definition / power to be used at the site ** --Find → Solve → Use ――How to create an analytical organization in less than 18 years? --There are four walls that are bigger than the problems of math and IT knowledge ―― 1. Maintaining motivation ―― 2. Cooperation with business divisions ―― 3. Contribution to management of the company ―― 4. Development of analytical organization members --Involve the business side in acquiring the budget for the sponsorship system --Rethink the number of team members --Create a place to discuss cost-effectiveness ――Think about how to incorporate it into your business

Fourth book: The power of analysis to change the company

Author: Kaoru Kawamoto
the publisher:Kodansha
Release Date: 2013/7/20

Impressions

――It was an old publication date, so I was suspicious, but it was a good book ――Chapter 1 is a topic that is often talked about (big data is not just big, etc.) --Explain the flow of analysis using a fictitious company as an example ――Working style as an analytical human resource ――Preparation for work --Professional, not specialist ――Not only familiar with the method but also familiar with the field ――I found from other books that analysis with methods is a problem, but I don't know the specific level ... ――In this book as well, if you find a "specialty field in business", you will be "expert level in the analysis of that field" and "know other analysis methods" ... ――When you think about the details, it feels like a contradiction.

• I do not understand.

Content introduction

--Same author as the above book --General misconceptions about data analysis, big data, etc. --"What is a data analysis specialist?" Working at a company —— Not only mastering mathematics, but also inspiring ways to utilize it --Responsible for results and analysis (review / interpret until convinced) —— Acquire a specialized field --Entering a place to make personal connections (society / community) --Disseminate information (society / community) --A wall to prevent unnecessary analysis

• data --Analysis
• KKD
• Cost-effectiveness

5th book: The power of data analysis A way of thinking that approaches causal relationships

Author: Koichiro Ito
the publisher:Kobunsha
Release Date: 2017/4/20

Impressions

――Since the author originally conducted empirical research on applied economics and policy, it is closer to empirical experiments. ――As the title says ** Can you analyze data, set issues, and evaluate results with "causality" in mind? A book called ** --A lot of space is devoted to "randomized trials" and "experimental design" --Easy to read. You can finish reading during long-distance travel while drawing a red line. -** It is not a book on how to utilize it for business **, but if you have knowledge of "demonstration experiments and causality" when thinking about "evaluation of results" and "setting of goals" I think I will be able to talk to the management ** -(It's a different story if you can listen without a feeling of refusal ...)

Content introduction

--There are many contents about how to measure "experiment" and "effect" ――Do not judge only by intuition and experience. The market may not react at all to the screen design that the web screen creator thought was good. ――If you can experiment on a large scale, randomize it and do an AB test before making a decision. ――It is difficult to measure the effect because the data of the “non-intervening group” cannot be obtained from the real data. --Visualize the data and create the result "I couldn't actually observe it, but it would have happened" and use it as the evaluation partner of the experimental results (** Potential results and parallel trend temporary / difference difference method * *)

Book 6: Introduction to Strategic Data Science-Concepts and Techniques for Business

Author: Foster Provost
Publisher: O'Reilly Japan
Release Date: 2014/7/19

Impressions

――This was long. About 500 pages ――The content feels a little old even from the year of publication ――Since it is a translation of an overseas author's book, it is difficult to read, which is peculiar to a translated book of a specialized book. -(There are many parts that are difficult to understand. I will survive with my own interpretation.) --The story you were looking for was in Chapters 11-14 and the appendix. -(Especially, I learned the necessary requirements for managers in the appendix and Chapter 13, environment maintenance and training, and project evaluation.) ――It is also important to have "love for your boss and subordinates", "communication skills with related departments", and "go outside and make strong acquaintances in a wide range of academic fields". ――Whether you are a manager or a management team, if you are going to do data science in a company, you need to know the analytical knowledge! Therefore, Chapters 2 to 6 and 10 introduce how to use the algorithm including theory.

Content introduction

--A book by an author who worked for a former major telephone company about how ** data science ** can be used in ** business ** to be successful for the company. ――Is data science in business "value?" From data? People in the data science field and managers with knowledge of data science ** "identify, collect data appropriately, evaluate achievements, and have a business impact. It is important to measure (including judging whether it can be measured) so that it can be deployed until it is used by people on the business side. "** ――For that purpose, in data analysis, after understanding the business, "clarify the issues", "determine whether the data is sufficient", and "data mining". --If a model can be created, it is necessary to perform "formulation, problem structuring, and modeling for problem solving" instead of Yoshi. ――Communicate with the business side and clarify the requirements as to how to develop it --Many citations are introduced ――It seems that data mining using in-house data is active overseas. ――Since the literature is just an introduction of examples and algorithms, it is recommended to read as much as time allows. ――It was an interesting document that you can measure "user's IQ" in "like articles" on facebook. - PNAS kosinski 2013

7th book: How to utilize the data that sleeps inside and outside the company-The point of view to find meaning in the data-

Author: Hayashi Hagikawa
the publisher:Sendenkaigi Co., Ltd.
Release Date: 2018/2/10

Impressions

――Since it is set in sales and marketing, the content is piercing for people in that field. -A good book that describes how beginners proceed with analysis to "speak with numbers" --There is a lot of extra content for those who need analysis that is not very relevant to surveys and targeting. ――Data analysis and hypothetical thinking are necessary for any analysis, so if you extract the necessary parts and format them, you can use them in other fields as well. ――I feel that it is a book for making business proposals from ** reports with in-depth analysis and interpretation ** rather than how to set issues.

Content introduction

――A novel-style book that focuses on "marketing" and makes "planning proposals" through "data analysis" --Collect facts for hypothetical thinking --Collect the missing facts from questionnaire surveys, interviews, and financial statements --Aggregate and visualize data and add interpretation to make it into a report --Narrow down customers from visualization and rate share analysis ――Based on the above, create a planning concept ―― 1. Target ―― 2. Needs ―― 3. Product ―― 4. Profit --Add issue analysis ―― 1. Feasibility ―― 2. Expected effect ―― 3. Business ――Finally, drop it into a format and make it a proposal —— Think about whether the plan and business model match

8th book: Introduction to the practical process of data analysis

Author: Anchibe
Publisher: Morikita Publishing
Release date: 2015/6/20

Impressions

――As mentioned in the subtitle, a book for "people who have no reliable boss or senior, no database or analysis know-how" --It is written in the order of purpose setting, data collection, preprocessing, analysis, operation (KPI), and measures. --The information you wanted was in "Purpose setting" and "Operation (KPI)" ――I was interested in the folder management of the author's analysis projects so that the coded ones would not be messed up. ――Although it is not divided into folders as finely as the author, it is good to manage scripts and data separately for each analysis project and analysis approach.

Content introduction

--Data has been collected. We analyzed and made a model with good accuracy. Is not of business value. --Set KPIs that can be judged as value (user retention rate? Billing amount? What is the solution to the problem) --Report the analysis results to the business side and come up with a plan on how to take measures and improvements. ――We will hold meetings during the measures and continue to interpret the results of the model and “utilize” them. --Check KPIs before and after measures and report whether they are improving ――If there are multiple related departments, understand which KPIs are beneficial for that department and show the necessary KPIs.

9th book: Practical data analytics course taught by Accenture professionals

Author: Takuya Kudo ・ Hogakusei
the publisher:Shoeisha
Release Date: 2016/5/30

Impressions

――The task setting for business utilization was in the first half chapter, but the volume is lighter than other books. ――As a whole, it's written so that it's too abstract and can be interpreted in any way. ――However, since it is also exhaustive, it is positioned as an introductory book with a wide range of data science coverage. ――Chapter 7 and later introduces methods and how to use packages that are hard to see using R. —— Feeling like reading “Data Science with R” --Market basket --A priori algorithm --Analysis and visualization using map data --Recommendation (collaborative filtering) --Hierarchical cluster analysis - Gephi,igraph ――Personally, it was very difficult to read. (Even a popular novelist feels that it is difficult to read sentences that do not suit him)

Content introduction

――About the method of improving the data infrastructure from the task setting ――I learned for the first time about the history of distributed processing technology and how it has evolved to make up for its disadvantages (Hadoop, MapReduce, DWH). --Chapter 7 and subsequent chapters introduce data analysis methods. ――It is necessary to set issues in order to create data science in business --There is a top-down / bottom-up approach to improve the analysis base

10th book: Introduction to data science learned from business use cases

Author: Ryuji Sakamaki, Yohei Sato
the publisher:SB creative
Release Date: 2014/6/24

Impressions

――There are few parts related to business utilization in Chapters 1 and 2, so it was out of the purpose of this time. --The business problem and R code are listed together, so it's a good practice book for those who just want to analyze it. --Business problem-> Find a plausible hypothesis-> Action plan-> Flow of analysis code using R --Unique because it uses data that requires preprocessing on purpose as an example --It is easy to imagine because it is an analysis example using structured (table) data that is extracted by SQL.

Content introduction

--Thinking about the gap between the ideal figure and the current situation --Phenomenon and problem are separated --Use analysis methods for various problems using virtual games as an example --Customer segmentation with clustering --Knowing the characteristics of continuous customers with a decision tree --Knowing the effect of advertising costs by multiple regression coefficient

11th: Revised 2nd Edition Data Scientist Training Reader [Acquire data analysis skills to become a professional! ]

Author: Multiple people are in charge of each chapter
the publisher:Technical Review Company
Release Date: 2016/8/25

Impressions

--Click here before revision

――The revised version is easier to read, so you should purchase the revised 2nd edition. ――In the first place, the color of business utilization is light, but if it is the first book for engineers, it is a pretty good book ――A book that you can see from this perspective when analyzing business data from the engineer side ――I hope that when you wear this analysis method and encounter data, you will get fluttering.

Content introduction

--A book that allows beginners involved in data science to know data science comprehensively --Data mining assuming business data (** Special Feature 2 **) is listed, so I made it a candidate. --Former Drecom, marketing analysis of Mr. Yohei Sato, who is also the author of the above "Introduction to Data Science" (There seems to be no script for official support? Handwritten sutras are difficult) --mixi Shimoda, Kimura's mining case --Hiroko Taisei's network analysis ――How do you connect to business? Although there are few talks, it is written that a cluster is created from customer data and what kind of marketing should be done to which layer is considered.

12th book: Learn while moving your hands Data mining that can be used in business

Author: Takashi Ozaki
the publisher:Technical Review Company
Release Date: 2014/8/21

Impressions

--Bloggers famous in the data science world and interesting blogs (those who work in Roppongi or Ginza) ――It may be the position of R's basic mining introduction book rather than for business utilization ――It may be good to try it with R and get an image, but --For a comprehensive R introductory book, I recommend other books (search based on words such as R, statistics, machine learning, etc.) ――In terms of business utilization, you may come up with the “seed of measures” from the analysis results, but it will stop there.

Content introduction

--Analysis with R after giving an example --Multiple regression of beer data --Determine the campaign effect Investigate the cause from the decision tree --Although the average value is different, analyze "not a fluke" considering statistics

13th book: Textbook for deep learning practice

Author: Japan Deep Learning Association
the publisher:Nikkei BP
Release Date: 2019/10/25

Impressions

――If you have knowledge about machine learning and deep learning, you can imagine what kind of technology solved the problem. --A book that delves into AI news articles ――What I wanted to know was more than an introduction to the technology ――How did you come up with the problem, what did you want to improve and how did you choose machine learning? --Verification work on the way and selection of models in it ――At what stage was the management decision? --Investing in projects and organizing ――Where did you throw the data scientist? ――It was, but these were not satisfied. --If you already have a data science team and are ready to promote it company-wide, you can use it when you say "** Read this book introduced in Japanese rather than thinking about diversion from an English paper **" so

Content introduction

――There are examples of deep learning used by each company. --G test recommended books (because the test raises problems with technology and its social implementation)

Finally, the summary I felt

――Rather than learning the analysis method, I learned the philosophy of "What is data analysis as a business person?" (I feel) ――You need to be prepared to proceed company-wide (if not, it's painful but bottom-up) ――Analyzing data by thinking about "how to use it for value" ――A sense of ownership that not only analyzes but also follows the user ――Hearing and using easy-to-use personal connections ――The result is "responsible at the same level as the site" ――Data scientists cannot be utilized due to miscellaneous tasks such as PoC cost reduction, self-reliance, useless PowerPoint creation & presentation ――After the analysis, in order to connect it to the measures, we will give a powerful explanation in "causal analysis" and squeeze it. ―― “Analysis results used” means selling to the site, winning the credit of the key man, persuasive explanation, polite explanation of how to use, creating an easy-to-read UI, analyzing the reason why it is no longer used and following up, and negotiating with the application infrastructure team. Modularize requirements definition and machine learning mechanism and outsource ... ** There are many things to do **

end

It was more difficult to get the author's mind than to read a book on analytical methods. In order to actually use the contents of the book, it is necessary to abstract and apply it according to the corporate culture.

~~ What is a data scientist? ~~