Time Series Analysis by JavaScript LL matsuri 2013
9 likes•3,388 views
This document discusses time series analysis and visualization using JavaScript. It introduces Series.js, a utility library for time series data that provides methods for aggregation, statistics, and visualization. Sample time series CPU and memory usage data is provided, and it is summarized using Series.js methods to aggregate the data by minute. Future work includes improving performance, documentation, and client-side capabilities of Series.js.
![Summarize
keyvalues = {}
for d in data
time = new Date(d.time.getFullYear(), d.time.getMonth(),
d.time.getDate(), d.time.getHours(), d.time.getMinutes())
keyvalues[time] = [] unless keyvalues[time]
keyvalues[time].push d
aggregated = []
for time, values of keyvalues
obj = {time: time, cpu: 0, memory: 0}
for d in values
obj.cpu += d.cpu
obj.memory += d.memory
aggregated.push obj
console.log aggregated
13](https://image.slidesharecdn.com/llmatsuri2013-130824105101-phpapp01/85/Time-Series-Analysis-by-JavaScript-LL-matsuri-2013-13-320.jpg)
![Series.js APIs example
# class methods
Series.sum([1,2,3,4,5]) #=> 15
Series.mean([1,2,3,4,5]) #=> 3
# instance methods
s = new Series()
s.put d for d in [1,2,3,4,5]
s.mean() #=> 3
18](https://image.slidesharecdn.com/llmatsuri2013-130824105101-phpapp01/85/Time-Series-Analysis-by-JavaScript-LL-matsuri-2013-18-320.jpg)
![Series.js APIs example
# accessor
Series.y((d)-> d.v).sum([{v:1},{v:2},{v:3},{v:4},{v:
5}]) #=> 15
s = new Series().y((d)-> d.v)
s.put d for d in [{v:1}, {v:2}, {v:3}, {v:4}, {v:5}]
s.mean() #=> 3
19](https://image.slidesharecdn.com/llmatsuri2013-130824105101-phpapp01/85/Time-Series-Analysis-by-JavaScript-LL-matsuri-2013-19-320.jpg)
![Series.js APIs example
# aggregation
data = [{t: 1, v:1}, {t: 1, v:2}, {t:2, v:3}, {t:2, v:4},
{t:1, v:5}]
Series.y((d)-> d.v).aggregation(Series.sum).key((d)-
> d.t)(data)
#=> [{t:1, v: 8}, {t: 2, v: 7}]
20](https://image.slidesharecdn.com/llmatsuri2013-130824105101-phpapp01/85/Time-Series-Analysis-by-JavaScript-LL-matsuri-2013-20-320.jpg)
Time Series Analysis by JavaScript LL matsuri 2013
- 1. Time Series analysis by JavaScript LL matsuri 2013/08/24 1
- 2. Time Series analysis by JavaScript 前処理 LL matsuri 2013/08/24 2
- 3. {twitter: “muddydixon”} ✓Data Mining ✓Data Visualization ✓Love D3.js ✓working @ NIFTY 3
- 5. Visualization ✓What is Visualization? communicating information clearly and effectively 5
- 6. Visualization on Web ✓Profit ✓Watch KGI / KPI continuously ✓Share visualization via URL ✓empowered representation by HTML5 / css3 -> D3.js 6
- 7. Visualization on Web : D3.js 7
- 8. Time Series Visualization ✓Visualize “time series” of ✓KGI / KP of services ✓Metrics of system resources 8
- 13. Summarize keyvalues = {} for d in data time = new Date(d.time.getFullYear(), d.time.getMonth(), d.time.getDate(), d.time.getHours(), d.time.getMinutes()) keyvalues[time] = [] unless keyvalues[time] keyvalues[time].push d aggregated = [] for time, values of keyvalues obj = {time: time, cpu: 0, memory: 0} for d in values obj.cpu += d.cpu obj.memory += d.memory aggregated.push obj console.log aggregated 13
- 14. Problems Large Time perspectives / Statistics Avoid to get data each time Hourly / Daily / Weekly / Monthly / Yearly Summary / Average / Stdev 14
- 16. Series.js 16
- 17. Series.js ✓Providing utility methods for time series ✓sum, sum square, mean, median, quantile ✓variance, standard devience, covariance, correlation ✓auto covariance, auto correlation ✓Providing aggregation ✓minutes, hour, day, week, month, year ✓Coming soon ✓auto regression, spectrum ✓moving average, holt-winters, exponential moving average 17
- 18. Series.js APIs example # class methods Series.sum([1,2,3,4,5]) #=> 15 Series.mean([1,2,3,4,5]) #=> 3 # instance methods s = new Series() s.put d for d in [1,2,3,4,5] s.mean() #=> 3 18
- 19. Series.js APIs example # accessor Series.y((d)-> d.v).sum([{v:1},{v:2},{v:3},{v:4},{v: 5}]) #=> 15 s = new Series().y((d)-> d.v) s.put d for d in [{v:1}, {v:2}, {v:3}, {v:4}, {v:5}] s.mean() #=> 3 19
- 20. Series.js APIs example # aggregation data = [{t: 1, v:1}, {t: 1, v:2}, {t:2, v:3}, {t:2, v:4}, {t:1, v:5}] Series.y((d)-> d.v).aggregation(Series.sum).key((d)- > d.t)(data) #=> [{t:1, v: 8}, {t: 2, v: 7}] 20
- 22. Summarize with Series.js Series.y((d)-> d.cpu).aggregation(Series.sum) .minute((d)-> d.time)(data) #=> [{ t: 'Sun Jan 13 2013 23:29:00 GMT+0900 (JST)', y: 2.0095 }, { t: 'Sun Jan 13 2013 23:30:00 GMT+0900 (JST)', y: 3.4370 }, { t: 'Sun Jan 13 2013 23:31:00 GMT+0900 (JST)', y: 1.58180 }, { t: 'Sun Jan 13 2013 23:32:00 GMT+0900 (JST)', y: 2.2742 }, { t: 'Sun Jan 13 2013 23:33:00 GMT+0900 (JST)', y: 3.4541 }, { t: 'Sun Jan 13 2013 23:34:00 GMT+0900 (JST)', y: 4.6035 }, { t: 'Sun Jan 13 2013 23:35:00 GMT+0900 (JST)', y: 2.1165 }, { t: 'Sun Jan 13 2013 23:36:00 GMT+0900 (JST)', y: 1.6177 }, 22
- 26. CROSS 2014 26