在R中叠加预测时间序列与原始序列及2017年1-4月预测需求

Hey there! Let's walk through how to get your 2017 Jan-Apr forecast results, compare them with your actual values, and plot both the original time series and forecast together in R. Here's the step-by-step solution:

First, let's adjust your original code to avoid variable overwrites and add steps for handling actual values and overlay plotting:

# Read the raw data
w_data <- read.csv("C:/Users/admin/Documents/aggrmonth.csv", sep=";", dec=",")

# Create time series object for historical data (2015-2016)
historical_ts <- ts(w_data$new, frequency = 12, start = c(2015, 1))

# Plot the historical time series
plot.ts(historical_ts, main = "Historical Monthly Data (2015-2016)", ylab = "Value")

# Load forecast package and fit Holt-Winters model
library(forecast)
hw_model <- HoltWinters(historical_ts)

# Forecast the next 4 months (2017 Jan-Apr)
forecast_result <- forecast(hw_model, h = 4)

# Store your actual values for 2017 Jan-Apr (replace the Apr value with your real data)
actual_2017 <- c(1020, 800, 1130, 950)
actual_ts <- ts(actual_2017, frequency = 12, start = c(2017, 1))

You can pull the forecast values directly from the forecast_result object. Here's how to view and extract them:

# View the full forecast output (including confidence intervals)
print(forecast_result)

# Extract just the numeric forecast values with month labels
forecast_values <- as.numeric(forecast_result$mean)
names(forecast_values) <- c("Jan-17", "Feb-17", "Mar-17", "Apr-17")

cat("\n2017 Jan-Apr Forecast Values:\n")
print(forecast_values)

To make a clear side-by-side comparison, create a summary data frame:

comparison_df <- data.frame(
  Month = c("Jan-17", "Feb-17", "Mar-17", "Apr-17"),
  Actual_Value = actual_2017,
  Forecast_Value = forecast_values,
  Absolute_Error = actual_2017 - forecast_values
)

print(comparison_df)

We have two straightforward options to visualize the overlay:

Option 1: Base R Plot (Simple & Lightweight) #

# Set plot limits to include 2017 data
plot.ts(historical_ts, 
        xlim = c(2015, 2017.3), 
        ylim = range(c(historical_ts, actual_2017, forecast_values)),
        main = "Historical Data + 2017 Forecast & Actuals", 
        ylab = "Value")

# Add forecast line and 95% confidence intervals
lines(forecast_result$mean, col = "blue", lwd = 2)
lines(forecast_result$lower[,2], col = "blue", lty = 2)
lines(forecast_result$upper[,2], col = "blue", lty = 2)

# Add actual 2017 values as points connected by lines
lines(actual_ts, col = "red", lwd = 2, type = "b")

# Add legend for clarity
legend("topleft", 
       legend = c("Historical", "Forecast", "95% CI", "2017 Actual"),
       col = c("black", "blue", "blue", "red"), 
       lty = c(1,1,2,1), 
       lwd = c(1,2,1,2))

Option 2: ggplot2-Based Plot (Polished & Customizable) #

Using the autoplot function from the forecast package gives a cleaner look:

library(ggplot2)

# Combine historical and 2017 actual data into one time series
full_actual_ts <- ts(c(historical_ts, actual_2017), frequency = 12, start = c(2015, 1))

# Generate the overlay plot
autoplot(full_actual_ts, series = "Actual Data") +
  autolayer(forecast_result$mean, series = "Forecast") +
  autolayer(forecast_result$lower[,2], series = "95% Lower CI", lty = 2) +
  autolayer(forecast_result$upper[,2], series = "95% Upper CI", lty = 2) +
  labs(title = "Historical Data + 2017 Jan-Apr Forecast & Actuals",
       y = "Value", x = "Year") +
  theme_minimal() +
  scale_color_manual(values = c("Actual Data" = "red", "Forecast" = "blue", 
                                "95% Lower CI" = "darkblue", "95% Upper CI" = "darkblue"))

内容的提问来源于stack exchange，提问作者psysky