A Probability Based Approach for Processing Dimension Missing Data

Processing missing value is one of the most important task in data mining. A great many applications, such as social commercial record, biological systems and remote sensing network, in which not only data values from particular features but even data dimension information may also be missing. Such missing values are known as dimension missing values-standard operation over these data may result in unrepresentable or uncertain problems. To tackle this problem of dealing with dimension missing data, in this paper, we first propose a probabilistic model to managing such data. Then, instead of enumerating all possible cases to recover the missed dimensions, we develop an effective and efficient bound confidence approach to speed up the retrieval process. A concrete evaluation using real data sets is reported, which shows that our method is effective and efficient on dimension incomplete data.