One-Step ahead Forecasting with Prophet in R

Question

I have built a custom function to do one-step-ahead forecasting using Prophet in R. But my code is not working don't know what is the problem. I have taken the rolling window size of 100 and trained the data with 60 days to forecast the next 1 entry in the data. But, my model is not working as it supposes to, please help me.

# loading the data:

s <- read.csv(url('https://ondemand.websol.barchart.com/getHistory.csv?apikey=c3122f072488a29c5279680b9a2cf88e&symbol=zs*1&type=dailyNearest&backAdjust=false&startDate=20100201'))

data <- s[c(3, 7)]

data <- as_tibble(data) %>% 
  mutate(tradingDay = as_date(tradingDay))

Here is my code:

# One-step ahead forecasting with Prophet

Prophet_prediction <- function(k) {
  
  range_validation <- 100
  n_ahead <- 1
  
  train_tbl <- data %>% slice((1 + k):(60 + k))
  valid_tbl <- data %>% slice((60 + 1 + k):(60 + k + range_validation)) 
  test_tbl  <- data %>% slice((60 + k + range_validation + 1):(60 + k + range_validation + n_ahead))
  
  train_df <- bind_rows(train_tbl, valid_tbl) %>% select(1:2)
  test_df <- test_tbl %>% select(1:2)
  
  # Prophet model: 
  my_prophet <- prophet(train_df[,2] %>% ts(start = 1))
  # Use the model for forecasting: 
  predict.prophet <- forecast(my_prophet, h = 1)$yhat %>% as.vector()
  
  actual_predicted_df_test <- test_df %>% 
    mutate(predicted = predict.prophet) 
  
  return(actual_predicted_df_test)

}

lapply(0:50, Prophet_prediction) ->> Prophet_prediction
Prophet_prediction <- do.call("bind_rows", Prophet_prediction)
view(Prophet_prediction)

But it gives me this error:

Error in attr(data, "tsp") <- c(start, end, frequency) : 
object is not a matrix

Answer 1

I can see at least a few errors. First, you are missing the closing bracket of your function. Second, you did not specify what k is. Additionally, the function has some issues.

First, this is wrong:

my_prophet <- prophet(train_df[,2] %>% ts(start = 1))

According to ?prophet, you need a

Dataframe containing the history. Must have columns ds (date type) and y, the time series.

This is also wrong

  predicted_prophet <- forecast(my_prophet, h = 1)$mean %>% as.vector()

Use function predict.prophet to predict using a prophet model. The object returned by predict.prophet does not have column mean. You might be looking for $yhat instead.