A Fine Slice Of SQL Server
Hadley Wickham reports on a Lubridate update:
Date time rounding (with
round_date(),floor_date()andceiling_date()) now supports unit multipliers, like “3 days” or “2 months”:ceiling_date(ymd_hms("2016-09-12 17:10:00"), unit = "5 minutes") #> [1] "2016-09-12 17:10:00 UTC"
If you handle date and time data in R, Lubridate is a tremendous asset.
Comments closedJaya Mathew has a SQL Server R Services template for predictive maintenance:
To illustrate the scenario, we will focus on companies who operate machines which encounter mechanical failures. These failures lead to downtime which has cost implications on any business, hence most companies are interested in predicting the failures ahead of time so that they can proactively prevent them. This scenario is aligned with an existing R Notebook published in the Cortana Intelligence Gallery but works with a larger dataset where we will focus on predicting component failures of a machine using raw telemetry, maintenance logs, previous errors/failures and additional information about the make/model of the machine. This scenario is widely applicable for almost any industry which uses machines that need maintenance. A quick overview of typical feature engineering techniques as well as how to build a model will be discussed below.
Understanding when machines are likely to break down is a very interesting statistical problem. Check out the template.
Comments closedMala Mahadevan discusses how to perform a Chi Square test:
For any dataset to lend itself to the Chi Square test it has to fit the following conditions –
1 Both variables are categorical (in this case – exposure to smoking – yes/no, and health condition – sick/not sick are both categorical).
2 Researchers used a random sample to collect data.
3 Researchers had an adequate sample size.Generally the sample size should be at least 100.
4 The number of respondents in each cell should be at least 5.
This is an easy case for using R over T-SQL—the Chi Square test is built in, whereas you have to roll your own T-SQL code. Mala does show you how to do this from within SQL Server R Services as well.
Comments closedWinston Chang reports that Shiny version 0.14 is now available:
If your Shiny app contains computations that take a long time to complete, a progress bar can improve the user experience by communicating how far along the computation is, and how much is left. Progress bars were added in Shiny 0.10.2. In Shiny 0.14, we’ve changed them to use the notifications system, which gives them a different look.
Important note: If you were already using progress bars and had customized them with your own CSS, you can add the
style = "old"argument to yourwithProgress()call (orProgress$new()). This will result in the same appearance as before. You can also callshinyOptions(progress.style = "old")in your app’s server function to make all progress indicators use the old styling.
It looks like they’ve made some good progress with Shiny.
Comments closedMala Mahadevan explains correlation coefficients:
The statistical definition of Pearson’s R Coefficient, as it is called, can be found in detail here for those interested. A value of 1 indicates that there is a strong positive correlation(the two variables in question increase together), 0 indicates no correlation between them, and -1 indicates a strong negative correlation (the two variables decrease together). But you rarely get a perfect -1, 0 or 1. Most values are fractional and interpreted as follows:
High correlation: .5 to 1.0 or -0.5 to 1.0.
Medium correlation: .3 to .5 or -0.3 to .5.
Low correlation: .1 to .3 or -0.1 to -0.3.
Mala includes R and T-SQL code so you can follow along.
Comments closedMilind Paradkar discusses clean data:
We decided to do a quick check and took a sample of 143 stocks listed on the National Stock Exchange of India Ltd (NSE). For these stocks, we downloaded the 1-minute intraday data for the period 1/08/2016 – 19/08/2016. The aim was to check whether Google finance captured every 1-minute bar during this period for each of the 143 stocks.
NSE’s trading session starts at 9:15 am and ends at 15:30 pm IST, thus comprising of 375 minutes. For 14 trading sessions, we should have 5250 data points for each of these stocks. We wrote a simple code in R to perform the check.
I like this post because it exposes a data quality issue people don’t tend to think about very often: when all of the data is legitimate and correctly-structured, but there are gaps in the available data set. This is one of many data quality problems you’ll run into, so it may be important to have a plan in place in case you hit this scenario.
Comments closedFang Zhou and Wee Hyong Tok have released a case study on a telephone company’s customer churn:
In the case of telco customer churn, we collected a combination of the call detail record data and customer profile data from a mobile carrier, and then followed the data science process — data exploration and visualization, data pre-processing and feature engineering, model training, scoring and evaluation — in order to achieve the churn prediction. With a churn indicator in the dataset taking value 1 when the customer is churned and taking value 0 when the customer is non-churned, we addressed the problem as a binary classification problem and tried varioustree-based models along with methods like bagging, random forests and boosting. Because the number of churned customers is much less than that of non-churned customers (making the data set quite unbalanced), SMOTE (Synthetic Minority Oversampling Technique) was applied to adjust the proportion of majority class over minority class in the training data set, thus further improving model performance, especially precision and recall.
All the above data science procedures could be implemented with base R. Rather than moving the data out from the database to an external machine running R, we instead run R scripts directly on SQL Server data by leveraging the in-database analytics capability provided by SQL Server R Services, taking advantage of the rich and powerful CRAN R packages plus the parallel external memory algorithms in the RevoScaleR library. In what follows, we will describe the specific R packages and algorithms that we used to implement the data science solution for predicting telco customer churn.
They have provided the relevant materials in GitHub as well.
Comments closedAllison Tharp looks at NFL play-by-play data using R:
Lets look at how teams played on offense depending on where they were on the field (their yardline) and the down they were on. The fields in our dataframe that we will care about here are yfog (yards from own goal), type (rush or pass), dwn (current down number: 1,2,3, or 4). We will want a table with each of these columns as well as a sum column. That way, we can see how many times a pass attempt was done on the 4th down when a team was X yards from their own goal.
To do this, we will use a package called plyr. The Internet says that this package makes it easy for us to split data, mess with it, and then put it back together. I am not convinced the tool is easy, but I haven’t spent too much time with it.
Check it out for some ideas on what you can do with R.
Comments closedDavid Kun introduces the R Database Layer:
It is important to note that the SQL statements generated in the background are not executed unless explicitly requested by the command as.data.frame. Hence, you can merge, filter and aggregate your dataset on the database side and load only the result set into memory for R.
In general the design principle behind RDBL is to keep the models as close as possible to the usual data.frame logic, including (as shown later in detail) commands like aggregate, referencing columns by the \($\) operator and features like logical indexing using the \([]\) operator.
Check it out. I’m not particularly excited about this for one simple reason: SQL is a better data retrieval and connection DSL than an R-based mapper. I get the value of sticking to one language as much as possible. I also get that not all queries need to be well-optimized—for example, you might be running queries on a local machine or against a slice of data which is not connected to an operational production environment. But I’m a big fan of using the right tool for the job, and the right tool for working with relational databases (and the “relational” part there is perhaps optional) is SQL.
Comments closedThe first satRdays event will take place in Budapest on September 3rd:
This is a very exciting project with great interest from the R and more general data science community — in the past short 2 months (since we opened registration for the conference):
More than 160 persons signed up and paid for attendance from 17 countries so far (around 50-50% mix of academic and industry tickets, 30-70% mix of foreign and Hungarian attendees)
We received almost 40 voluntary talk proposals in a few weeks of time while the CfP was open
25 selected & awesome speakers agreed on to present at the conference
I’d like to see this take off, similar to SQL Saturdays.
Comments closed