gistfile1.txt

dfnew%>%filter(!is.na(comp3_provided))%>%
  dplyr::select(ogn_id,curve_day,
                feed_provided,
                comp1_provided,
                comp2_provided,comp3_provided)

### Forecast summed feed curves over time
testdata2<-dfnew%>%
  group_by(date_interval, lcn_id, newField) %>%
  mutate(comp2_provided   = ifelse(is.na(comp2_provided),0,comp2_provided),
         comp1_provided_2 = ifelse(curve_day<30, comp1_provided, NA), 
         comp2_provided_2 = ifelse(curve_day<30, comp2_provided, NA))%>%
  dplyr::select(ogn_id,
                lcn_id,
                curve_day,
                feed_provided,
                comp1_provided,
                comp1_provided_2,
                comp2_provided,
                comp2_provided_2,
                newField,
                date_interval,
                animals_actual,
                feed_provided_cum)
testdata2$pred1<-predict(m1_gam_comp1$gam, 
                       newdata=testdata2)
testdata2$pred2<-predict(m1_gam_comp2$gam, 
                         newdata=testdata2)
g1<-testdata2%>%
  group_by(ogn_id, curve_day)%>%
  summarize(prov1    = sum(comp1_provided_2),
            pred1    = sum(pred1))%>%
  ggplot(., aes(x=curve_day))+
  geom_point(aes(y=prov1), col="black")+
  geom_line(aes(y=prov1),  col="black")+
  geom_point(aes(y=pred1), col="red", alpha=0.5)+
  geom_line(aes(y=pred1),  col="red", alpha=0.5)+
  theme_bw()+
  labs(x="Day in Curve", 
       y="Feed", 
       fill="Farm", 
       title="Forecasting for Component 1 per Farm ", 
       subtitle ="Black is actual provided - Red is forecasted")+
  facet_wrap(~ogn_id)

g2<-testdata2%>%
  group_by(ogn_id, curve_day)%>%
  summarize(prov2    = sum(comp2_provided_2),
            pred2    = sum(pred2))%>%
  ggplot(., aes(x=curve_day))+
  geom_point(aes(y=prov2), col="black")+
  geom_line(aes(y=prov2),  col="black")+
  geom_point(aes(y=pred2), col="seagreen", alpha=0.5)+
  geom_line(aes(y=pred2),  col="seagreen", alpha=0.5)+
  theme_bw()+
  labs(x="Day in Curve", 
       y="Feed", 
       fill="Farm", 
       title="Forecasting for Component 2 per Farm ", 
       subtitle ="Black is actual provided - Green is forecasted")+
  facet_wrap(~ogn_id)

### CUMULATIVE OVER TIME IS NOT CORRECT!! 
g3<-testdata2%>% 
  arrange(ogn_id, lcn_id, newField, curve_day)%>%
  group_by(ogn_id,curve_day)%>%
  summarize(prov1        = comp1_provided_2,
            prov1_sum    = sum(comp1_provided_2),
            pred1        = pred1,
            pred1_sum    = sum(pred1),
            prov2        = comp2_provided_2,
            prov2_sum    = sum(comp2_provided_2),
            pred2        = pred2,
            pred2_sum    = sum(pred2))%>%
  filter(row_number()==1)%>%
  group_by(ogn_id)%>%
  summarize(curve_day = curve_day,
            prov1_cumsum = cumsum(prov1_sum),
            pred1_cumsum = cumsum(pred1_sum),
            prov2_cumsum = cumsum(prov2_sum),
            pred2_cumsum = cumsum(pred2_sum))%>%
  distinct()%>%
  ggplot(., aes(x=curve_day))+
  geom_point(aes(y=prov1_cumsum ),col="black", alpha=0.5)+
  geom_line(aes(y=prov1_cumsum ), col="black", alpha=0.5)+
  geom_point(aes(y=pred1_cumsum), col="red", alpha=0.5)+
  geom_line(aes(y=pred1_cumsum),  col="red", alpha=0.5)+
  geom_point(aes(y=prov2_cumsum ),col="black", alpha=0.5)+
  geom_line(aes(y=prov2_cumsum ), col="black", alpha=0.5)+
  geom_point(aes(y=pred2_cumsum), col="seagreen", alpha=0.5)+
  geom_line(aes(y=pred2_cumsum),  col="seagreen", alpha=0.5)+
  theme_bw()+
  labs(x="Day in Curve", 
       y="Cumulative Feed", 
       fill="Farm", 
       title="Forecasting for Component 1 & 2 per Farm",
       subtitle="Black is actual - red(comp1) & green(comp2) are forecasted from day 30 onwards")+
  facet_wrap(~ogn_id)
grid.arrange(g1,g2,g3,ncol=1)