5 inbound probability density function : 無料・フリー素材/写真
5 inbound probability density function / ewedistrict
| ライセンス | クリエイティブ・コモンズ 表示-継承 2.1 |
|---|---|
| 説明 | This is the inferred probability density function for the arrival of a 5 inbound bus at Fillmore & McCallister during weekdays.A little detail on my model and parameter estimation: there is a normal probability density function (PDF) corresponding to each stop, with a bias and scale shared by all stop PDFs. The PDF of any given vehicle arriving is simply the superposition of all individual stop PDFs. Let f(x) be the probability density function of any vehicle arriving f(x). The log likelihood of a number of independent stop times is sum([log(f(time)) for time in times]) to use the python notation. We're interested in normal distribution parameters that maximize this log likelihood computed against our body of actual stop times. I rigged up a function and subjected it to scipy.optimize.minimize, and it found a likelihood maxima at bias=99.76 and scale=217.30. That is to say, vehicles are late on average of 99.76 seconds with a standard deviation of 217.30 seconds. In the middle of the day it all blurs together into a constant PDF.Next up: checking if this PDF is well-calibrated. |
| 撮影日 | 2012-11-13 15:06:34 |
| 撮影者 | ewedistrict , Boston, US |
| 撮影地 |