An algorithm explanation for calculating content heat

Define the heat of a content

What kind of content is popular? This is actually a relative concept, in the face of different needs, it is expressed differently.

$H = \frac{W + I}{(T+1)^{G}}$

Among them:

• HHH: The heat of content
• WWW: numerical expression of content quality, for example, the quality of an article can be defined as the number of likes of the article, or as others according to the needs of the product
• III: The initial quality of the new content, the specific meaning of which will be understood in subsequent explanations
• Gravity, or the speed at which a piece of content becomes less popular, the greater the Gravity, the faster a piece of content refreshes
• TTT: time

You can see that HHH has four variables in this formula H(W,I,T,G)H(W,I,T,G)H(W,I,T,G), which is the same formula as Hacker News and Reddit, except that the logic used to define the quality of the content WWW is different.

1) heat
$H$With the time
$T$The relationship between

First, let’s look at a relatively simple relationship between heat and time:

• WWW: 100
• III: 10
• GGG: 1.5
• TTT: time

$100 + H = \ frac {10} {(T + 1) ^ 1.5} {}$

Clearly, the more time, the less heat.

2) heat
$H$And initial mass
$I$, time,
$T$The relationship between

• WWW: 0
• III: [2, 4, 6]
• GGG: 1.5
• TTT: time

We can see that for different initial mass III, the initial heat of the content is different, and in subsequent time decay, if other values remain constant, the higher the initial mass, the greater the heat of the content.

3) the heat
$H$And initial mass
$I$, gravity
$G$, time,
$T$The relationship between

• WWW: 0
• III: [2, 4, 6]
• GGG: [1.5, 1.2, 1.8]
• TTT: time

Even if the initial mass is high, the heat drops off quickly when the gravity is high.

4) the heat
$H$With the quality
$W$Initial mass
$I$, gravity
$G$, time,
$T$The relationship between

• WWW: [90, 190, 90]
• III: 10
• GGG: [1.5, 1.5, 1.8]
• TTT: time

We see three intersections:

• AAA: W=90W=90W=90, G= 1.5g = 1.5g =1.5 and I=10I=10I=10 intersection

Explanation: A content mass of 90 is less hot than a brand new content with an initial heat of 10 at the AAA intersection, about 3.64 units of time later.

• BBB: W=190W=190W=190, G= 1.5g = 1.5g =1.5 and I=10I=10I=10 intersection
• CCC: W=90W=90W=90, G= 1.8g = 1.8g =1.8 and I=10I=10I=10 intersection

Explanation: A content mass of 190 is less popular than any new content after 6.36 units of time. A content mass of 90, with a gravity of 1.8, is less hot than any new content only after 2.59 units of time.

We can see the following conclusions:

1. The higher the initial quality III, the faster the popularity of older content is overtaken by the latest production
2. The higher the quality of a piece of content, the longer it remains popular than new content
3. The higher the gravity GGG is, the faster the content heat decays, and the faster the heat of the old content is surpassed by the newly produced content

Thus, at a given initial mass III and gravity GGG, the WWW that a content can acquire determines its popularity and the time it takes to be overtaken by new content.

5) Quality value
$W$With the time
$T$The assumption of a linear positive correlation

Assuming that the numerical expression of content quality is positively correlated with time, for example, we express the quality of an article with the number of likes, then within a limited time, the longer the time, the higher the number of likes:

$W(T) = c\times{T}$

And CCC stands for how fast a new content gets quality WWW, which is how fast a new content gets likes, so

• WWW: c * * * \ times {T} c * T
• III: 10
• GGG: 1.5
• TTT: time

$H = \ frac {\ times c + 10} {T} {(T + 1) ^ 1.5} {}$

We tested three CCC’s: 20, 25 and 30

We see three intersections:

• AAA: W=20×TW=20\times{T}W=20×T and I=10I=10I=10
• BBB: W=25×TW=25\times{T}W=25×T and I=10I=10I=10
• CCC: W=30×TW=30\times{T}W=30×T and I=10I=10I=10

Explanation: The higher the CCC, the higher the quality of a piece of content per unit of time, which means the content itself is more popular. For intersection AAA, the content will be overtaken by the heat of a new content, even if it maintains a mass growth rate of 20×T20\times{T}20×T after 1.62 units of time. As can be seen from the intersection of CCC, if its WWW growth rate is 30×T30\times{T}30×T, it will be surpassed by the heat of new content after 6.46 units of time.

6) Quality value
$W$With the time
$T$The logarithmic positive correlation hypothesis

Of course, one problem with the above hypothesis is that the quality expression of an article is difficult to have a long-term positive correlation with time, that is to say, the longer the article exists, the lower the number of likes it will get later. Therefore, we can assume that WWW and TTT are log-dependent:

$W(T) = c\times{\log_{d}{(T+1)}}$

• CCC: How fast you get likes
• DDD: Get the coefficient at which the like speed starts to decay over time

We see three intersections:

• AAA: W = 20 x log ⁡ (T + 1) 2 W = 20 \ times {\ log_ {2} {(T + 1)}} W = 20 * log2 (T + 1) and 10 I = I = 10 I = 10 intersection
• BBB: W = 20 x log ⁡ 1.5 W (T + 1) = 20 \ times {\ log_ 1.5} {{(T + 1)}} W = 20 x log1.5 (T + 1) and 10 I = I = 10 I = 10 intersection
• CCC: W=20×TW=20\times{T}W=20×T and I=10I=10I=10

Explanation: Comparing AAA to CCC, we can see that AAA content is overtaken faster by new content because the quality number increases less over time.

Content on the use of heat in nuggets

$H = \frac{W + I}{(T+1)^{G}}$

• 
  $H$: The heat of a contentrankIndexArticle
• 
  $W$: that is, of a contenthotIndexArticleIs a weighted sum of the number of views, comments, and likes on an article
• 
  $I$: is the initial value of a contentrankIndexUserIs the influence factor of the article author
  • The influence factor is related to the author’s own historical excavation value
• GGG: A decaying gravity parameter
• TTT: The length of the article since it was published

As a gold digging community, we want to bring more good content to our readers. Numerically, we want the total WWW to be high, while balancing it with TTT, the length of time between traffic to new content and traffic to a good content.

• CCC: The quality expression of an article itself, how quickly it gets likes and gets read, i.e. click through rate and like rate
• DDD: Whether the like rate and click rate will decay with the time of the article, by how much

Speaking in human terms: how can a story with 1,000 likes, 10,000 views, and 100 comments be outdone by the popularity of a new article? The heat calculation of the whole community is a dynamic optimization problem.

Parameter adjustment under different problems

• Too much new content, not enough recent hits
  1. Lower the initial value III
  2. Reduced attenuated gravity GGG
  3. Improve content quality by weighting the number CCC
• Content is hot, new things don’t see
  1. Increase the initial value III
  2. Increase attenuation gravity GGG
  3. Reduced weighted CCC
• One moment it’s an old hot one, the next it’s a particularly new one
  1. When GGG is small, III is too large