The cosine measure normalizes document length so that long
documents are not favored simply because they have more terms.
Later work empirically showed that, for the TREC collection,
longer documents actually have a higher probability of being
relevant [1]. We show that the slope should be recalibrated for fundamentally
different document collections. We recomputed the normalization
for the web track and obtained a 16% improvement in our
baseline run. The resulting effectiveness from this sole
modification was an average precision higher than any other
group who submitted results to the TREC-10 web track.
Design Database Process (review)
The process includes:
· Gathering user or business's requirement
· Develop E-R Model bases user or business's requirements
· Convert E-R Model go to relationship gatherings (table)
· Relationship normalization to remove anomaly
· Implementation goes to database by make table for each relationship already most normalization
Database normalization
* Normalization is the establishment of the database structure process so that most of the ambiguity can be removed.
* Normalization stage, starting from the most mild (1NF) to most stringent (5NF)
* Usually only up to the level of 3NF or BCNF because already sufficient to generate the table-a table of good quality.
* Why do normalization?
Ø Optimizing table structures
Ø Increase speed
Ø To remove income of same data.
Ø More efficient in the use of storage media
Ø Reduce redundancy
Ø Avoid anomalies (insertion anomalies, deletion anomalies, update anomalies).
Ø Improved data integrity
* A table saying good (efficient) or if the normal 3 to meet the following criteria :
Ø If there is decomposition (decomposition) table, it must be guaranteed safe decomposition it (Lossless-Join Decomposition). That is, after the table is described / decompositioned a new table-table, the table-table can generate a new table with the same exact.
Ø Maintain dependence on the functional changes in data (Dependency preservation).
Ø Does not violate Boyce-Code Normal Form (BCNF)
* If the third criteria (BCNF) can not be met, then at least the table does not violate the Normal Form of the third stage (3rd Normal Form / 3NF).
Information Systems Analysis and Design [Part-6]
Posting by my name is onix on 2:01 AM
DATABASE NORMALIZATION
Design Database Process (review)
The process includes:
· Gathering user or business's requirement
· Develop E-R Model bases user or business's requirements
· Convert E-R Model go to relationship gatherings (table)
· Relationship normalization to remove anomaly
· Implementation goes to database by make table for each relationship already most normalization
Database Normalization
Normalization is process of database structure forming so mainly part of ambiguity can be removed. Normalization phase is begun from lightest phase (1NF) until tightest (5NF). Usually, only coming up with level 3NF or BCNF because has adequate enough to yield the tables of which with quality goodness.
Why done by normalization?
• Optimizing table structures
• Increasing speed
• Removing same data inclusion
• More efficient in storage media purpose
• Reducing redundancy
• Avoiding anomaly (insertion anomalies, deletion anomalies, and anomaly's update).
• Increased data integrity
One table is said well (efficient) or normal if accomplish 3 criterions as follows:
- If there is decomposition (parsing) table, therefore the decomposition shall be secured safe (Lossless Join Decomposition). It’s mean, after that table is untied / at decomposition becomes new tables, that new tables can result original table equally exactly
- Its preserve dependency functional at the moment data change (Dependency Preservation).
- Don't breach Boyce-Code Normal Form (BCNF)
If the third criterion (BCNF) can't be accomplished, therefore at least that table not breach the third Normal Form (3rd Normal Form/ 3NF).
Functional Dependency
Functional Dependency figuring relationship attributes in one relationship. An attribute is said functionally dependant on the other if we utilize the attributes value to determine the other attribute value. Symbol that is utilized is for representing functional dependency.
Read functionally determines.
Notation: A B
A and B is attribute of one table. It means that functionally A determines B or B depend on A, if and only if available 2 rows data with same value of A, therefore value of B also same.
Notation:

Functional Dependency:
* NRP -> Name
* Study_lecture, NRP -> grade/value
Non Functional Dependency:
* Study_lecture -> NRP
* NRP -> grade/value
* Functional Dependency from the value table :
Ø NRP -> Name
Because for each value Nrp the same, then the value of the same name.
Ø {Study_lecture, NRP } -> grade/value
Because the value of attributes depending on the NRP and Study_lecture together. In another sense Study_lecture for the NRP and the same, they also rated the same, because Study_lecture and the NRP is a key (is unique).
Ø Study_lecture ->/ NRP
Ø NRP ->/ value
First Normal Form - 1NF
A table on the form said to be normal if I did not reside in the unnormalized form of a table, where there is a kind of field multiplication and field that allows a null (empty).
not be allowed :
* Attributes which many valuable (Multivalued attributes).
* Attributes composite or a combination of both.
So that :
* Price of the attribute domain must be atomic rates.
v Example Student Data as follows:

The tables above are ineligible for 1NF. Decomposition becomes:
Collegian table:

Hobby table:

Second Normal Form - 2NF
* Normal form 2NF met in a table if it meets the form of 1NF, and all the attributes than the primary key, have a full Functional Dependency on primary key
* A table does not meet 2NF, if there are attributes that dependence it (Functional Dependency) are only partial (only depending on the part of the primary key)
* If there are attributes that have no dependence on the primary key, then the attributes must be moved or removed
* Functional dependence X -> Y , said full if it is said to delete an attribute A from X means that Y is no longer dependent functional.
* Functional dependence X -> Y , said partial if a delete attribute A from X means that Y is functionally dependent.
* Relation scheme R in the form 2NF if every non-primary key attribute A Î R depend on the full functional primary key R.
* For example :
Ø The following table meet 1NF, but not including 2NF

Ø Does not meet 2NF, because (NIM, CodeMk) is regarded as the primary key whereas:
{NIM, CodeMk} à NameMhs
{NIM, CodeMk} à Address
{NIM, CodeMk} à LectureStudy
{NIM, CodeMk} à Sks
{NIM, CodeMk} à LetterValue
Ø The table needs to decomposition to become some table eligible 2NF.
Ø Functional dependency it as follows:
o {NIM, CodeMk} à LetterValue (fd1)
o NIM à {NameMhs, Address} (fd2)
o CodeMk à {LectureStudy, Sks} (fd3)
Ø So that :
o fd1 (NIM, CodeMk, LetterValue) à Table of Value
o fd2 (NIM, NameMhs, Address) à Table of Student
o fd3 (CodeMk, LectureStudy, Sks) à Table of LectureStudy
Third Normal Form - 3NF
* Normal form 3NF fulfilled if the form meets 2NF, and if there are no non-primary key attribute that has a dependence on non-primary key attributes of the other (transitive dependencies).
* For example :
Ø This following table of students eligible 2NF, but does not meet to 3NF :
Ø Because there are non-primary key attribute (ie, City and Provincial), which has a dependence on non-primary key attributes of the other (ie PostCode):
PostCode à {City, Province}
Ø So that the table need to decomposition to becomes
Ø Student (NIM, NameMhs, Street, PostCode)
o PostCode (PostCode, Province, City)
Boyce-Codd Normal Form (BNCF)
* Boyce-Codd Normal Form constraint has a stronger form of the third Normal. To be BNCF, relations must be in the form of first Normal and forced each of the attributes depends on the function in the super key attributes.
* In the example below there is a relationship seminar, Primary Key is NPM + Seminar.
Students may take one or two seminars. Each seminar requires 2 each of the students and led by one of the 2 seminar. Each leader can only take one seminar course. NPM in this example and the Seminar show a Preceptor.
Seminar Relationship

Seminar Relationship form is third Normal Form, but not BCNF since Seminar Code is still dependent function on Counselor, if each Counselor gets to teach just one seminar. Dependent seminar on one attribute is not super key as presupposed by BCNF. Therefore Seminar relationship shall be broken down as two which is:

Seminar - Preceptor Relationship

Fourth and fifth Normal Form
Relationship in fourth normal form (4 NF) if relationship in BCNF and not contains of multi value dependency. To remove multi value dependency from one relationship, we are divide relationship become two new relationships. Each relationship contains of two attributes that have multi value relationship.
Relationship in fifth normal form (5NF) get business with property is called join without marks sense information loss (lossless join). The fifth normal Form (5 NF also called PJNF (projection join normal form). This case is very rare to appearance and hard to be detected practically
References:
1. Agus Sanjaya ER,S.Kom, M.Kom, slide presentation : Database and ER-Diagram
2. Network Intrusion Detection: Evasion,
Traffic Normalization, and End-to-End Protocol Semantics