Comparing Go to Python/R

Image credit: R logo, Gopher

Evaluating Go

Unlike Python/R, Go is a compiled language that is more verbose but is said to run faster. For example, Uber, Amex and KhanAcademy find benefits with Go.

To benchmark Go’s performance and runtime against Python and/or R, here are various cases:

1. Performing least squares regression of the Anscombe Quartet (1973)
2. Computing summary statistics of the California Housing Prices (Miller 2015)
3. Web crawling and scraping of Wikipedia
4. Identifying anomalies in the MNIST dataset

Least squares regression with Python and R

The Go implementation is benchmarked for runtime with a previous implementation by Miller (2015) in Python/R as a reference. Least squares regression is implemented in Go using the stats package. Python was significantly slower compared to R and Go implementations with runtimes of 1.36s, 0.04s, 0.173s for Python, R and Go. While R was less verbose and a bit faster than Go, Go’s testing package ensured identitcal least squares coefficients of 0.5 and 3 for each Anscombe dataset during development.

See my Github repository for further details.

Summary statistics with Python and R

The Go implementation is done using the stats package. Runtimes are compared using ’time’ before commands in the command line to compare with Python’s pandas.describe() and R’s summary functions. The operations are run 100 times for each implementation. Python was significantly faster compared to R and Go implementations with ‘real’ runtimes of 2.27s, 4.10s, 5.02s for Python, R and Go respectively. While Python and R were less verbose and a bit faster than Go, Go’s testing package ensured similar summary statistics as Python for each of the seven variables (value, income, age, rooms, bedrooms, pop, hh) during development.

To run locally, download or git clone this project:

1
2
3
4
5

git clone https://github.com/asaraog/msds431week4.git
cd msds431week4
time ./Week4
time python3 runHouses.py 
time  Rscript runHouses.R

See my Github repository for further details.

Web crawling and scraping with Python

This project implemented web crawling and scraping of Wikipedia (described by Chanda 2021) webpages in Go using Colly. The Go implementation is benchmarked for runtime using ’time’ before commands in the command line to compare with Python’s implementation for the same 10 webpages using scrapy. Python was significantly slower compared Go implementations with ‘real’ runtimes of 15.9s and 0.6s for Python and Go respectively. While Python was less verbose than Go, Go is more scalable and has concurrency support to allow for even faster processing using Colly.

To run locally, download or git clone this project:

1
2
3
4
5
6

git clone https://github.com/asaraog/msds431week5.git
cd msds431week5
time ./Week5

cd WebFocusedCrawlWorkV001
time python3 run-articles-spider.py

See my Github repository for further details.

Identifying anomalies using isolation forests with Python and R

Isolation forests are used as an unsupervised learning method to identify anomalies or outliers. It was introduced by Liu 2008 by observing that path lengths for anomalies were significantly SHORTER by averaging over many trees. He introduces an anomaly score to normalize comparisons with HIGHER scores indicating more abnormality. Hyperparameters were kept the same across languages with 1000 trees and 256 samples. The go-iforest package was used for the analysis in Go. Comparison code for R and Python was adapted from Miller 2023. Runtimes were significantly lower in Go with runtimes of 5.66s, 19.02s, and 1m 42.87s for Go, R and Python respectively. This is likely due to the goroutines utilized in the go-iforest package.

To run locally, download or git clone this project:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10

git clone https://github.com/asaraog/msds431week7.git

cd msds431week7/go/whole
time ./Week7

cd ../../python
time python3 isolationForest.py

cd ../r
time Rscript isotreeForest.R

See my Github repository for further details.

Conclusion

Programs written using Go had similar performances to Python/R in basic statistical learning applications such as least squares regression and summary statistics. Efficient results were obtained utilizing test-driven development. In applications with more scalability, Go performed very well. For both web scraping/crawling and identifying anomalies, Go performs more than 10 times faster than Python. This could be attributed to Go’s concurrency support with goroutines.

References

Anscombe, F. J. 1973. “Graphs in Statistical Analysis.” The American Statistician 27 (1): 17–21. https://doi.org/10.2307/2682899.

Chanda, Subha. 2021. “Web Scraping with Go.” ScrapingBee. 2021. https://www.scrapingbee.com/blog/web-scraping-go/.

Liu, Fei Tony, Kai Ming Ting, and Zhi-Hua Zhou. 2008. “Isolation Forest.”. In ICDM ‘08: Proceedings of the 2008 Eighth IEEE International Conference on Data Mining, December 2008, 413–422.

Miller, Tom. “Testing Go for Statistics,”. MSDS 431: Data Engineering with Go. Course at Northwestern University, Chicago, IL, June 19, 2023.

Miller, Tom. “Data Cleaning, Frames and Pipelines,”. MSDS 431: Data Engineering with Go. Course at Northwestern University, Chicago, IL, June 19, 2023. https://github.com/ThomasWMiller/jump-start-mnist-iforest

Miller, Tom. “Command-Line Applications,”. MSDS 431: Data Engineering with Go. Course at Northwestern University, Chicago, IL, July 1, 2023.

Miller, Tom. “Crawling and Scraping the Web,”. MSDS 431: Data Engineering with Go. Course at Northwestern University, Chicago, IL, July 2, 2023.