Part 1 of the Exploratory Data Analysis can be found here : https://rpubs.com/stanspwan/275198
In this Part 2 , I try the prophet forecasting package from facebook on the speeding violations data set.
hchart(tseries, name = "test") %>%
hc_add_theme(hc_theme_economist()) %>%
hc_credits(enabled = TRUE, text = "Data Source : Chicago Police Department", style = list(fontSize = "13px")) %>%
hc_title(text = "Times Series plot of speed camera violations in Chicago 2014-2016") %>%
hc_legend(enabled = TRUE)
my_model <- prophet(df)
## Initial log joint probability = -4.28681
## Optimization terminated normally:
## Convergence detected: relative gradient magnitude is below tolerance
# creating time stamps for two years
future <- make_future_dataframe(my_model, periods = 365 * 2)
#Quick sneak peek
head(future)
## ds
## 1 2014-07-01
## 2 2014-07-02
## 3 2014-07-03
## 4 2014-07-04
## 5 2014-07-05
## 6 2014-07-06
tail(future)
## ds
## 1640 2018-12-26
## 1641 2018-12-27
## 1642 2018-12-28
## 1643 2018-12-29
## 1644 2018-12-30
## 1645 2018-12-31
## 2. Forecast prediction
forecast <- predict(my_model, future)
tail(forecast[c('ds', 'yhat', 'yhat_lower', 'yhat_upper')])
## ds yhat yhat_lower yhat_upper
## 1640 2018-12-26 4.627034 4.412295 4.841228
## 1641 2018-12-27 4.618168 4.406873 4.835586
## 1642 2018-12-28 4.578050 4.366928 4.788702
## 1643 2018-12-29 4.155719 3.934655 4.376804
## 1644 2018-12-30 4.151697 3.932252 4.357084
## 1645 2018-12-31 4.585387 4.370674 4.786704
## 3. Plotting the forecast
plot(my_model, forecast)
# Weekly, Monthly and Annual breakdown
prophet_plot_components(my_model, forecast)
Disclaimer/Note: This may not be the most scientific forecasting, I am just familiarizing myself with Facebookâs Prophet forecasting package and trying to use it.Any suggestions are welcome.