wwydmanski commited on
Commit
63c62b0
1 Parent(s): db7d4c9

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +92 -0
README.md ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ task_categories:
3
+ - tabular-regression
4
+ - tabular-classification
5
+ tags:
6
+ - tabular
7
+ size_categories:
8
+ - 10K<n<100K
9
+ ---
10
+
11
+ ## Source
12
+ Source: [UCI](https://archive.ics.uci.edu/ml/datasets/BlogFeedback)
13
+
14
+ ## Data Set Information:
15
+
16
+ This data originates from blog posts. The raw HTML-documents
17
+ of the blog posts were crawled and processed.
18
+ The prediction task associated with the data is the prediction
19
+ of the number of comments in the upcoming 24 hours. In order
20
+ to simulate this situation, we choose a basetime (in the past)
21
+ and select the blog posts that were published at most
22
+ 72 hours before the selected base date/time. Then, we calculate
23
+ all the features of the selected blog posts from the information
24
+ that was available at the basetime, therefore each instance
25
+ corresponds to a blog post. The target is the number of
26
+ comments that the blog post received in the next 24 hours
27
+ relative to the basetime.
28
+
29
+ In the train data, the basetimes were in the years
30
+ 2010 and 2011. In the test data the basetimes were
31
+ in February and March 2012. This simulates the real-world
32
+ situtation in which training data from the past is available
33
+ to predict events in the future.
34
+
35
+ The train data was generated from different basetimes that may
36
+ temporally overlap. Therefore, if you simply split the train
37
+ into disjoint partitions, the underlying time intervals may
38
+ overlap. Therefore, the you should use the provided, temporally
39
+ disjoint train and test splits in order to ensure that the
40
+ evaluation is fair.
41
+
42
+ ## Attribute Information:
43
+
44
+ 1...50:Average, standard deviation, min, max and median of them attributes 51...60 for the source of the current blog post. With source we mean the blog on which the post appeared.
45
+ For example, myblog.blog.org would be the source of the post myblog.blog.org/post_2010_09_10
46
+
47
+ 51: Total number of comments before basetime
48
+
49
+ 52: Number of comments in the last 24 hours before the
50
+ basetime
51
+
52
+ 53: Let T1 denote the datetime 48 hours before basetime,
53
+ Let T2 denote the datetime 24 hours before basetime.
54
+ This attribute is the number of comments in the time period
55
+ between T1 and T2
56
+
57
+ 54: Number of comments in the first 24 hours after the
58
+ publication of the blog post, but before basetime
59
+
60
+ 55: The difference of Attribute 52 and Attribute 53
61
+
62
+ 56...60:
63
+ The same features as the attributes 51...55, but
64
+ features 56...60 refer to the number of links (trackbacks),
65
+ while features 51...55 refer to the number of comments.
66
+
67
+ 61: The length of time between the publication of the blog post
68
+ and basetime
69
+
70
+ 62: The length of the blog post
71
+
72
+ 63...262:
73
+ The 200 bag of words features for 200 frequent words of the
74
+ text of the blog post
75
+
76
+ 263...269: binary indicator features (0 or 1) for the weekday
77
+ (Monday...Sunday) of the basetime
78
+
79
+ 270...276: binary indicator features (0 or 1) for the weekday
80
+ (Monday...Sunday) of the date of publication of the blog
81
+ post
82
+
83
+ 277: Number of parent pages: we consider a blog post P as a
84
+ parent of blog post B, if B is a reply (trackback) to
85
+ blog post P.
86
+
87
+ 278...280:
88
+ Minimum, maximum, average number of comments that the
89
+ parents received
90
+
91
+ 281: The target: the number of comments in the next 24 hours
92
+ (relative to basetime)