class: center, middle, inverse, title-slide # Data Science Preparation ### <a href="http://www.rui-wang.com/">Rui Wang</a> ### 2018/03/24 --- # Outline ### SQL ### Probability ### Statistics ### Coding ### Machine Learning ### Case Studies --- # SQL -- - PRIMARY KEY * Uniquely identify each record in a database table * Cannot contain NULL values * Only one PRIMARY KEY -- - FOREIGN KEY * Used to link two tables together * Can contain NULL values * Refer to the PRIMARY KEY -- - Structure ``` select ... from ... join ... on ... where ... group by ... having ... order by ... ``` --- # SQL -- - [Intro To Database](https://lagunita.stanford.edu/courses/Engineering/db/2014_1/about) * [Pratice Questions](https://github.com/wangruinju/SQL_Resources/blob/master/Stanford%20SQL%20practice/SQL%20exercise.Rmd) -- - Online Platforms * [Leetcode DataBase](https://leetcode.com/problemset/database/) * [Hackerrank](https://www.hackerrank.com/) * [Vertabelo Acedemy](https://academy.vertabelo.com/) -- - [Analytical Cases](https://community.modeanalytics.com/sql/tutorial/sql-business-analytics-training/) * Invertigating a Drop in User Engagement * Understanding Search Functionality * Validating A/B Test Results -- - Reference * [W3Schools](https://www.w3schools.com/SQl/default.asp) --- # SQL - Content Type table: content_id | content_type (comment/ post) | target_id * If it is comment, target_id is the user_id who posts it. * If it is post, then target_id is NULL. What is the distribution of comments? -- ``` select cnt, count(cnt) as freq from (select content_id, count(target_id) cnt from table group by content_id) a group by cnt; ``` --- --- # Probability -- - [Statistical Inference](https://www.amazon.com/Statistical-Inference-George-Casella/dp/0534243126) * Probability and Basic Statistics: Chapter 1-5 * Hypothesis Testing: Chapter 6-12 -- - [MIT: Introduction to Probability and Statistics](https://ocw.mit.edu/courses/mathematics/18-05-introduction-to-probability-and-statistics-spring-2014/) * [Youtube Video List](https://www.youtube.com/playlist?list=PLUl4u3cNGP61Oq3tWYp6V_F-5jb5L2iHb) -- - [A Green Book: A Practical Guide to Quantitative Finance Interview](https://www.amazon.com/Practical-Guide-Quantitative-Finance-Interviews/dp/1438236662) * Data Science: Chapter 4 -- - [Cheat Sheet](https://static1.squarespace.com/static/54bf3241e4b0f0d81bf7ff36/t/55e9494fe4b011aed10e48e5/1441352015658/probability_cheatsheet.pdf) * Cover Most of Knowledge Points --- # Probability -- - Cards What is probability of getting one pair of card from a deck of 52 cards? -- 3/51 -- What is probability of getting two cards in the same suit from a deck of 52 cards? -- 12/51 -- What’s the probability of getting two cards that are not in the same suit and not a pair from a deck of 52 cards? -- 36/51 -- - Coins You randomly draw a coin from 100 coins - 1 unfair coin (head-head), 99 fair coins (head-tail) and roll it 10 times. If the result is 10 heads, whats the probability that the coin is unfair? -- `\(\frac{1/100}{1/100*1 + 99/100*(1/2)^{10}} = \frac{1024}{1033}\)` --- # Probability - Seattle Raining You’re about to get on a plane to Seattle. You want to know if you should bring an umbrella. You call 3 random friends of yours who live there and ask each independently if it’s raining. Each of your friends has a 2/3 chance of telling you the truth and a 1/3 chance of messing with you by lying. All 3 friends tell you that “Yes” it is raining. What is the probability that it’s actually raining in Seattle? -- p denotes the probablity of raining at Seattle, then we will have the conditional probablity as `$$\frac{(2/3)^3*p}{(2/3)^3*p + (1/3)^3*(1-p)}$$` -- If `\(p = 1/4\)`, we have the answer as 8/11. --- # Probability - Birthday There are 30 people in a class. What is the probablity of at least two people who have the same birthday? -- Think the opposite the questions about the probablity of all the people who have different birthdays. `\(P = 1 - \frac{365*364*...*336}{365^{30}}\)` What is the probabily of that there are exactly two people who have the same birthday? (similar to card problems like suit and pair) -- Choose one pair out of 30 people and assign one out of 365 days to their birthday. Then consider the probablity of the rest 28 people who have different birthdays. `\(P = \frac{\binom{30}{2}*365*364*...*337}{365^{30}}\)` --- # Statistics - Linear regression - Assumptions: linearity, independence, homogeneity and normality - Inference and interpretation - Issues like outlier and multicollinearity - Logistics regression - Invalid assumptions compared with linear regression - Interpretation - Hypothesis testing - p value, `\(H_0\)`, `\(H_1\)`, type I error, type II error, power - Confidence intervals - Online course - [Biostatistics Module](http://sphweb.bumc.bu.edu/otlt/mph-modules/menu/) --- # Coding - Data Wrangling Using Pandas - Basics - Selection - Missing values - String - Merge - Groupby - Plot - Time series - Data in/out - Resources - [Python for Data Analysis](https://github.com/wesm/pydata-book) - [Pandas Documentation](https://pandas.pydata.org/) - [Pandas Exercises](https://github.com/guipsamora/pandas_exercises) --- # Coding - Data Structure Using Python - Python * [Problem Solving with Algorithms and Data Structures using Python](http://interactivepython.org/runestone/static/pythonds/index.html) * [Google Python Course](https://developers.google.com/edu/python/) * [MIT: Introduction to Algorithms](https://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-006-introduction-to-algorithms-spring-2008/) * [Youtube Video List](https://www.youtube.com/playlist?list=PLUl4u3cNGP61Oq3tWYp6V_F-5jb5L2iHb) -- - [Leetcode](https://leetcode.com/problemset/algorithms/) * [My Python Solution](https://github.com/wangruinju/Rui_Python_Leetcode) -- .pull-left[ * Part I * Linked List * Binary Search * Two Points * Bit Manipulation * Math * Stack * Hash Table] .pull-right[ * Part II * Backtrack * Tree * DFS * BFS * String * Array * Dynamic Programming] --- # Machine Learning - Basics - Supervise learning vs unsupervise learning - Linear regression - Logistics regression and Linear Discriminat Analysis - Cross validation and bootstraping - Regularization: l1/l2 - Tree models: CART/Random Forest/Gradient Boosting Tree - SVM and kernels - Clustering: k-means, gaussian mixture model and hierarchical methods - PCA for data reductation - Resources - [An Introduction to Statistical Learning](http://www-bcf.usc.edu/~gareth/ISL/ISLR%20First%20Printing.pdf) - [The Elements of Statistical Learning](https://web.stanford.edu/~hastie/Papers/ESLII.pdf) - [Andrew Ng: Maching Learning on Coursera](https://www.coursera.org/learn/machine-learning/home/welcome) and [My solution](https://github.com/wangruinju/Machine-Learning-Coursera) - [CS229: Maching Learning](http://cs229.stanford.edu/) - [Hunag-yi Lee videos](https://www.youtube.com/playlist?list=PLJV_el3uVTsPy9oCRY30oBPNLCo89yu49) - [Other great videos](https://www.youtube.com/playlist?list=PLaXDtXvwY-oDvedS3f4HW0b4KxqpJ_imw) --- # Case Studies - Structure - Define the business problems - Make plan for the analysis - Segment data - Summarize the insights - Make business decisions - Methods - Segment analysis - Trend analysis - Funnel analysis - User behavior analysis - Retention analysis - AB test: sample size/effect size/significant level/power/potential issues - Resources - [A/B Testing on Udacity](https://www.udacity.com/course/ab-testing--ud257) - [Design of Experiments](https://onlinecourses.science.psu.edu/stat503/node/1) - [A/B testing articles](https://engineering.linkedin.com/blog/topic/ab-testing) --- # All of these are part of my preparation for data science interviews. I was encouraged by a lot of my friends at the beginning and now I am willing to introduce my limited experience to people who are in need. Cheers!