Assignment 3 : Profet model
Hetal Sawant
2/15/2022
Time Series Dataset : Beijing Air Quality
Forecating Air Quality with Profet
changping<-read.csv("C:/Users/Hetal Sawant/Desktop/Spring sem/Forecasting/HW/PRSA_Data_Changping_20130301-20170228.csv")
attach(changping)
#str(changping)
head(changping)## No year month day hour PM2.5 PM10 SO2 NO2 CO O3 TEMP PRES DEWP RAIN wd
## 1 1 2013 3 1 0 3 6 13 7 300 85 -2.3 1020.8 -19.7 0 E
## 2 2 2013 3 1 1 3 3 6 6 300 85 -2.5 1021.3 -19.0 0 ENE
## 3 3 2013 3 1 2 3 3 22 13 400 74 -3.0 1021.3 -19.9 0 ENE
## 4 4 2013 3 1 3 3 6 12 8 300 81 -3.6 1021.8 -19.1 0 NNE
## 5 5 2013 3 1 4 3 3 14 8 300 81 -3.5 1022.3 -19.4 0 N
## 6 6 2013 3 1 5 3 3 10 17 400 71 -4.5 1022.6 -19.5 0 NNW
## WSPM station
## 1 0.5 Changping
## 2 0.7 Changping
## 3 0.2 Changping
## 4 1.0 Changping
## 5 2.1 Changping
## 6 1.7 Changping## No year month day hour PM2.5 PM10 SO2 NO2 CO O3 TEMP PRES DEWP RAIN wd
## 1 1 2013 3 1 0 3 6 13 7 300 85 -2.3 1020.8 -19.7 0 E
## 2 2 2013 3 1 1 3 3 6 6 300 85 -2.5 1021.3 -19.0 0 ENE
## 3 3 2013 3 1 2 3 3 22 13 400 74 -3.0 1021.3 -19.9 0 ENE
## 4 4 2013 3 1 3 3 6 12 8 300 81 -3.6 1021.8 -19.1 0 NNE
## 5 5 2013 3 1 4 3 3 14 8 300 81 -3.5 1022.3 -19.4 0 N
## 6 6 2013 3 1 5 3 3 10 17 400 71 -4.5 1022.6 -19.5 0 NNW
## WSPM station y1
## 1 0.5 Changping 2013-03-01
## 2 0.7 Changping 2013-03-01
## 3 0.2 Changping 2013-03-01
## 4 1.0 Changping 2013-03-01
## 5 2.1 Changping 2013-03-01
## 6 1.7 Changping 2013-03-01Assesing the Changping SO2 levels over the 5 years of period using Facebook Prophet model:
Estimating Changepoints
We can observe the change-points at the beginning (Jan- Feb) and end (Nov - Dec) of each year.
Observing Trends and Seasonality
As observed below, we can see that SO2 levels drop around Wednesday in weekly seasonality. While considering monthly seasonality we can see that SO2 levels drop between April to October.
We also observed SO2 levels impacting around New year’s day (1st Jan) and Chinese new year (1st Feb).
Assesing saturation points:
It is expected to have positive values for SO2 levels. Maximum SO2 levels observed are 340 and minimum is 10. We can see the same reflecting in below graphs. We also observed predicted values slightly towards negative number.
##Seasonality : With Additive and Multiplicative
Cross-validation to assess performance of the model
## y ds yhat yhat_lower yhat_upper cutoff
## 1 27 2015-03-25 4.291868 -20.30234 30.58157 2015-03-24
## 2 27 2015-03-25 4.291868 -22.46475 29.84741 2015-03-24
## 3 27 2015-03-25 4.291868 -21.35638 30.16130 2015-03-24
## 4 27 2015-03-25 4.291868 -20.81661 30.32415 2015-03-24
## 5 27 2015-03-25 4.291868 -20.82622 29.80059 2015-03-24
## 6 27 2015-03-25 4.291868 -22.92249 27.22851 2015-03-24## [1] "2015-03-24 GMT" "2015-04-23 GMT" "2015-05-23 GMT" "2015-06-22 GMT"
## [5] "2015-07-22 GMT" "2015-08-21 GMT" "2015-09-20 GMT" "2015-10-20 GMT"
## <ggproto object: Class ScaleDiscrete, Scale, gg>
## aesthetics: colour
## axis_order: function
## break_info: function
## break_positions: function
## breaks: waiver
## call: call
## clone: function
## dimension: function
## drop: TRUE
## expand: waiver
## get_breaks: function
## get_breaks_minor: function
## get_labels: function
## get_limits: function
## guide: legend
## is_discrete: function
## is_empty: function
## labels: waiver
## limits: NULL
## make_sec_title: function
## make_title: function
## map: function
## map_df: function
## n.breaks.cache: NULL
## na.translate: TRUE
## na.value: grey50
## name: Cutoff
## palette: function
## palette.cache: NULL
## position: left
## range: <ggproto object: Class RangeDiscrete, Range, gg>
## range: NULL
## reset: function
## train: function
## super: <ggproto object: Class RangeDiscrete, Range, gg>
## rescale: function
## reset: function
## scale_name: hue
## train: function
## train_df: function
## transform: function
## transform_df: function
## super: <ggproto object: Class ScaleDiscrete, Scale, gg>
