In many real-world data applications the available data is sparse and incomplete, and the data that is available can contain outliers or faulty measurements. For the purpose of further data processing, this can be viewed as a problem of missing values in the data. Missing values in the data, when not handled with care, can lead to multiple problems in data information and prediction applications:

• There may be insufficient (remaining) data to allow meaningful processing and provide useful information
• Missing values may be incorrectly interpreted during training of prediction models, leading to learning of incorrect patterns
• Missing values may be incorrectly interpreted during online prediction, leading to the use of incorrect input information in the forecasts

All three problem can lead to insufficient or unreliable information and predictions.

However, if data contains missing values, imputation can be used to estimate the missing values. Imputation fills in the missing values based on what can be expected given surrounding values in the time series or given similar situations in the data. Adapticon's DDM methodology has unique methods for imputation of missing data which are particularly well suited to various data sources. They are based on two types of imputation functions:

• Imputation by prediction
• Imputation by interpolation