Rice Production Forecast Using ARIMA Model

Introduction

• Forecasting the production of rice using ARIMA model
• Bivariate analysis
• Dataset: https://www.kaggle.com/code/vaibhavkhande/analysis-on-crop-production-in-different-countries/notebook
• Preprocess the data
  • Removed columns not required
  • replace() and dropna() to remove null values
  • Used groupby() to calculate the mean crop yield for each country

Plotting Rolling Statistics

• To understand trends and patterns [upward or downward trend]
• To detect changes in variance [shift in the underlying process (depending on SD changes)]
• To assess stationarity [statistical properties do not change over time]
• To improve forecasting accuracy [smooth out the data]

Augmented Dickey-Fuller (ADF) test

• To check whether the time series is stationary or not

Coupling AD-Fuller check with MA and moving SD

• To identify the trend and volatility of the data

  

Data Transformation

Logarithmic Transformation

• To stabilize the variance of the data
• To convert multiplicative relationships into additive relationships
• To make the data more normally distributed

Removing Trend With Moving Average

• To identify seasonal patterns
• To stabilize the variance of the data
• To make the time series stationary

Exponential Decay Transformation

• To stabilize the variance of the data
• To model and analyze the data using traditional time series techniques

Time Shift Transformation

• To compare and analyze data at different lags
• To identify the time delay between the two variables
• To better understand the relationship between them

Data Visualization

Line Chart

• To visualize crop prodction from 1990 to 2020

Histogram

• To find the frequency

Decompose

• To separate a time series into trend, seasonality, and residual (or noise)

Auto Correlation Function and Partial Auto Correlation

• ACF: the correlation of the time series with its past values over different lags
• PACF: the correlation of the time series with its past values over different lags, but without the influence of other lags

Auto Regression (AR) Model

• Uses past values of a variable to predict its future values
• Parameters
  • p, the number of lagged values (order of autoregression)
  • q, the number of past errors q (order of moving average)
  • d, the number of times the time series is differenced to make it stationary (order of differencing)

Moving Average Model

• Used to identify trends and patterns in data
• Used to make predictions about future trends

ARIMA (Auto Regression Integrated Moving Average)

• Used for time series analysis and forecasting 

Model Performance

Mean Square Errors

• To compare the predictions of the model against the actual data
• Calculated as the average of the squared forecast error values

Prediction and Reverse Transformation

• Prediction: the process of using a model to make an educated guess about an outcome or result
• Reverse transformation: the process of converting a transformed value back to its original form

Convert to Cumulative

• To transform the data so that each data point represents the cumulative sum of the previous data points

Inverse of Log is Exp

• To transform a logarithmic transformation back to its original scale

  

Forecast

• To visualize the forecasted values of an ARIMA model

Conclusion

• By implementing ARIMA Model predictive values are almost accurate to the actual outcomes
• Though the prediction from ARIMA model nearly is accurate, if there is a major breakdown in climatic conditions this model can not predict the accurate values