# Computational Intelligence for Missing Data Imputation, Estimation and Management Knowledge Optimization Techniques pdf

0
684

## Computational Intelligence for Missing Data Imputation, Estimation and Management Knowledge Optimization Techniques by Tshilidzi Marwala pdf

Computational Intelligence for Missing Data Imputation, Estimation and Management Knowledge Optimization Techniques by Tshilidzi Marwala pdf free download. In real life, a set of data invariably contains missing data. The problem then is to reconstitute the most probable values through processes such as interpolation and extrapolation before using that set. Methods for resolving the problem of missing data have been extensively explored in statistical texts (Abdella, 2005; Little & Rubin, 1987).

### Computational Intelligence for Missing Data Imputation, Estimation and Management Knowledge Optimization Techniques by Tshilidzi Marwala pdf

The initial work on compensating for missing data was focused on improving survey data. In this book, missing data interpolation is called imputation to distinguish it from the statistical approach. Imputation is viewed as an alternative approach to deal with missing data. There are two ways to deal with missing data: these are either to estimate the missing data or to delete any vector (data set) with missing value(s). This book focuses on methods that estimate the missing values. Of particular importance to the area of missing data interpolation is to analyze the nature of the missing data, and this is termed the missing data mechanism.

### Computational Intelligence for Missing Data Imputation, Estimation and Management Knowledge Optimization Techniques by Tshilidzi Marwala pdf

Little and Rubin (1987) categorized three missing data mechanisms, namely: Missing At Random (MAR), Missing Completely At Random (MCAR) and a non-ignorable case also known as Missing Not At Random (MNAR). In the first case, MAR occurs when the probability that variable X is missing depends on other variables, but not on X itself. An example of this is the case where two variables: the vibration level of a machine and its temperature, X are measured. If a very high vibration level causes the temperature sensor to fall off and thus high and subsequently low values of X become missing because of the other variable vibration level, this is termed MAR.

#### Computational Intelligence for Missing Data Imputation, Estimation and Management Knowledge Optimization Techniques by Tshilidzi Marwala pdf

MCAR occurs when the probability that variable X is missing is unrelated to the value of X itself or to any other variable in the data set. This refers to data sets where the absence of data does not depend on the variable of interest or of any other variable in the data set (Rubin, 1978). MNAR occurs when the probability of variable X missing is related to the value of X itself even if the other variables are controlled in the analysis (Allison, 2000). An example of this is when in a survey of weights of candidates, a person omits mentioning his or her weight because its value is very high. In analyzing survey data, these mechanisms are very powerful and useful. Knowing these mechanisms assists one in choosing which missing data imputation method is best to use.

##### Computational Intelligence for Missing Data Imputation, Estimation and Management Knowledge Optimization Techniques by Tshilidzi Marwala pdf

However, in many engineering problems, where on-line decision support tools are becoming widely used, these mechanisms are proving to be insignificant (Marwala & Hunt, 1999). For example, if an aircraft is flying over the Atlantic Ocean and one of its critical sensors fails, there is simply no time to investigate why that particular sensor has failed and, thereby, indentify its missing value mechanism. What ought to be done in this situation is to quickly estimate the sensorâ€™s value, so that an on-line autopilot system can continue to operate. In using decision support tools, if data become missing, it is extremely important, particularly for critical applications, that the missing data estimation technique is accurate.