Information Technology Reference
In-Depth Information
By our method, we clarify the relationship between each heterogeneous thing, event
and phenomenon.
The person harnesses the past experience and does prediction and prevention.
Our system can do the same things like this. That is, we can predict and prevention
of Big data. Therefore, we realize the one of the methods for decision mining.
The contributions of this paper are as follows:
•
We summarize essences of Big data analytics.
•
We propose a new concept for discovery, an anteroposterior correlation.
•
We represent a discovery method of anteroposterior correlation between each
heterogeneous thing, event and phenomenon.
This paper is organized as follows. In section 2, we survey the related works of
our proposed method. In section 3, we introduce our ethic of essences for Big data
analytics. After that, we propose our method, a discovery method of anteroposte-
rior correlation in section 4. We present some experiment results and discussion in
section 5. Finally, we conclude in section 6.
2
Related Works
Correlation or similarity measures for a discovery of relationships have been studied
for a long time. The most popular and basic method is vector space model [1].
Dimensionality reduction techniques of vector space model have been used for
developing traditional vector space models such as latent semantic indexing [2] and
the mathematical model of meaning [3, 4]. These techniques are applied to infor-
mation resources, characterized by elements in a flat domain. However, it is to be
noted that when the elements have a hierarchical structure, all the elements are not
orthogonal to each other. A few studies have used computational measures of direc-
tionality relationships [5] in an orthogonal vector space. The mathematical model of
meaning realizes a context-driven dynamic semantic computation. However, it has
to prepare a space for the semantic commutation before. Our method processes a
dynamic data-driven space creation corresponding to a context. That is, our method
does not have to prepare the space before. It is very an important difference, because
we cannot create the space or any schemas before in the open assumption. Currently,
we are in Big data era. In the Big data environment, we can aggregate various and
a lot of fragmental data. We cannot predict what kinds of the data we obtain in ad-
vance. Actually, the rise of the key-value store means that the schema cannot be
designed in advance. Since the data updates are faster and faster, we should change
dynamically also space for the semantic computations and analysis.
There have been studies defining similarity metrics for a discovery of relation-
ships, such as WordNet [6]. Rada et al. [7] have proposed a ”conceptual distance”
that indicates the similarity between concepts of semantic nets by using path lengths.
Some studies [8] [9] have extended and used the conceptual distance for information
retrieval. Resnik [10] has proposed an alternative similarity measure based on the
concept of information content. Ganesan et al. [11] have presented new similarity
Search WWH ::
Custom Search