A time series is a sequence of data points recorded at specific time points - most often in regular time intervals (seconds, hours, days, months etc.). Every organization generates a high volume of data every single day – be it sales figure, revenue, traffic, or operating cost. Time series data mining can generate valuable information for long-term business decisions, yet they are underutilized in most organizations. Below is a list of few possible ways to take advantage of time series datasets:
- Trend analysis: Just plotting data against time can generate very powerful insights. One very basic use of time-series data is just understanding temporal pattern/trend in what is being measured. In businesses it can even give an early indication on the overall direction of a typical business cycle.
- Outlier/anomaly detection: An outlier in a temporal dataset represents an anomaly. Whether desired (e.g. profit margin) or not (e.g. cost), outliers detected in a dataset can help prevent unintended consequences.
- Examining shocks/unexpected variation: Time-series data can identify variations (expected or unexpected) and abnormalities, detect signals in the noise.
- Association analysis: By plotting bivariate/multivariate temporal data it is easy (just visually) to identify associations between any two features (e.g. profit vs sales). This association may or may not imply causation, but this is a good starting point in selecting input features that impact output variables in more advanced statistical analysis.
- Forecasting: Forecasting future values using historical data is a common methodological approach – from simple extrapolation to sophisticated stochastic methods such as ARIMA.
- Predictive analytics: Advanced statistical analysis such as panel data models (fixed and random effects models) rely heavily on multi-variate longitudinal datasets. These types of analysis help in business forecasts, identify explanatory variables, or simply help understand associations between features in a dataset.