R语言分解2018年日度销售时间序列报错求助

Hey there! Let's get your sales forecasting script working properly. The error you're seeing happens because the decompose() and stl() functions expect a ts (time series) object with an explicitly defined frequency—right now you're passing a zoo object, which these functions can't interpret to find repeating cycles (like your weekly sales pattern).

What's Causing the Error? #

• decompose() and stl() rely on knowing the frequency of your time series (how often the pattern repeats). For daily data with weekly cycles, the frequency should be 7 (one week = 7 days).
• Your myzoo object is a zoo object, not a ts object. These decomposition functions don't work with zoo inputs directly.

Fixed Code Walkthrough #

Here's the adjusted version of your script that resolves the error and keeps your forecasting workflow intact:

# Load required packages
library(forecast)
library(sweep)
library(timetk)
library(zoo)

# Read the Data
# First, read raw values and convert to a numeric vector
sales_data <- scan("R Project/Dataset/MyData.txt")
# Skip the leading "X" value from your data file
sales_values <- sales_data[-1]

# Create a ts (time series) object with daily frequency (7 days/week)
# Start date: 2018-01-01, which is the 1st day of 2018
my_ts <- ts(sales_values, start = c(2018, 1), frequency = 7)

# Inspect and plot the time series
print(my_ts)
plot(my_ts, xaxt = "n")
# Add formatted date axis
date_inds <- seq(as.Date("2018-01-01"), as.Date("2018-12-31"), by = "day")
Axis(date_inds, side = 1, at = seq(date_inds[1], tail(date_inds, 1) + 60, by = "1 months"), format = "%b %Y")

# Decompose the time series (now works!)
composition <- decompose(my_ts)
plot(composition) # Plot decomposition components

# STL decomposition (also works now)
stl_result <- stl(my_ts, s.window = "periodic")
plot(stl_result)

# ARIMA forecasting
fit <- auto.arima(my_ts)
fore <- forecast(fit, h = 15, level = c(99.5))

# Plot the forecast with custom axis
plot(fore, xaxt = "n")
Axis(date_inds, side = 1, at = seq(date_inds[1], tail(date_inds, 1) + 60, by = "1 months"), format = "%b %Y")

Key Changes Made: #

1. Data Loading: Used scan() to read raw numeric values, then removed the leading "X" from your data file.
2. ts Object Creation: Replaced the zoo object with a ts object, specifying frequency = 7 to tell R your data has a weekly cycle. This is what fixes the decomposition error.
3. Decomposition Functions: Now decompose() and stl() can correctly identify the weekly pattern in your data, since they have the frequency information they need.

Quick Note on Zero Values #

Since your 0s represent holidays/Sundays, you might want to consider handling them (e.g., replacing with NA or interpolating) if you want more accurate decomposition/forecasting—right now, those zeros will pull down the seasonal component estimates. But that's an optional next step once your basic workflow is working!

内容的提问来源于stack exchange，提问作者Joan