In this section, I begin with the excel file of sales data, which I obtained from the Tableau Community Forum. As a recall, the data contains mostly categorical variables and components of the vectors from the description column. The index column is a timeseries format. The major objective of this section is to understand the general trends in the data, and gain some quick insights, and then predict and forcast the Sales of the category "Technology" within a given certain period of time of the Superstore data.The statistical significance of these observations will be also tested in 'Exploratory Data Analysis'.
The following performances has been done in this time series data:
- Data Exploration
- Checking for Seasonality
- Weekly and monthly resampling of sales
- Checking for trends
- Checking for stationarity using Rolling plots, Summary statistics, and statistical tests.
- Reduce noise by Aggragarion, Smoothing, and Polynomial Fitting techniques
- Smoothing: Moving Average and Exponentially weighted moving average.
- Decomposition of a time series data into its components: trend, seasonality, and noise. Prediction, Forecast, Validation of forecasts, and the dynamic Forecasts.
- Applying ARIMA model to forecast with the best values of p,d,and q (determined by correlation graphs and hyperparameter optimizations like AIC method).
- Accuracy of predictions
- Future improvements