Predicting Click-Through Rates of New Advertisements Based on the Bayesian Network

Algorithm 1

KBN construction.

Input:  𝑄={𝑞₁,𝑞₂,…,𝑞_𝑙}, a set of query;

    𝐾_𝐴={𝐾₁,𝐾₂,…,𝐾_𝑚}, a set of the known ads’ keywords;

    𝜀, threshold of the similarity of two keywords (0≤𝜀≤1)

Output:  𝐺=(𝑉,𝐸), DAG of the KBN

Variable:  𝑛, the number of nodes in the KBN

Steps:

    (1)    𝐾_𝑥←𝑄∩𝐾_𝐴

    (2)   𝑉←𝐾_𝑥,𝐸←{}

    (3)   For 𝑖←1 to 𝑛−1 Do

    (4)   For 𝑗←𝑖+1 to 𝑛 Do

    (5)      If sim(𝑘_𝑖,𝑘_𝑗)>𝜀 Then // By (2)

    (6)      If 𝑃(𝑘_𝑖∣𝑘_𝑗)≥𝑃(𝑘_𝑗∣𝑘_𝑖) Then 𝐸←𝐸∪{𝑘𝑗→𝑘_𝑖} // By (3)

    (7)      Else 𝐸←𝐸∪{𝑘_𝑖→𝑘_𝑗}

    (8)      End If

    (9)      End If

    (10) End For

    (11)  End For

    (12) Return 𝐺