CN111416741A - Event hotspot prediction method based on Internet technology - Google Patents

Abstract

The invention belongs to a prediction method, and particularly relates to an event hotspot prediction method based on an internet technology. It comprises the following steps: step 1: calculating the real-time attention of a single event; accumulating the real-time attention degrees of all the networks at the same time; step 2: attention of all events; respectively calculating the attention degrees of all events needing attention in the network; and step 3: calculating a short-term attention degree trend; calculating a short-term attention trend of the event; and 4, step 4: predicting; predicting the next stage condition; and 5: sorting; and (4) sequencing the results obtained in the step (4), wherein the sequencing result is the sequencing situation of the event hot spots. The invention has the following remarkable effects: the calculation complexity is low, and the economical efficiency and the effectiveness are good, so that the practical requirements are met. According to the method and the device, the influence degree of time on the event is fully considered, the influence degree of the past event development trend on the future is also fully considered, and the prediction is accurate.

Description

Event hotspot prediction method based on Internet technology

Technical Field

The invention belongs to a prediction method, and particularly relates to an event hotspot prediction method based on an internet technology.

Background

The management and control of network information are important components of internet information security. Good network information management and control can reduce negative information statement and highlight positive energy. One of the characteristics of the network information is that hot spots are generated irregularly, the burstiness is strong, some information is strong in persistence, and some information is quickly faded. If the network hotspot information can be predicted, the trend of the future hotspot information can be predicted in advance, so that manual intervention can be performed in advance.

The existing prediction method generally adopts modes such as an artificial application network, a Bayesian algorithm and the like to predict. These prediction methods are either computationally intensive or require a sufficient number of training samples, often with an economic or time cost that is not satisfactory. Therefore, a prediction method with simple implementation and fast calculation is needed.

Disclosure of Invention

Aiming at the defects of the prior art, the invention provides an event hotspot prediction method based on the Internet technology.

The invention is realized by the following steps: an event hotspot prediction method based on the Internet technology comprises the following steps:

step 1: calculating real-time attention of single event

Accumulating the real-time attention degrees of all the networks at the same time;

step 2: attention to all events

Respectively calculating the attention degrees of all events needing attention in the network;

and step 3: calculating short-term attention trends

Calculating a short-term attention trend of the event;

and 4, step 4: prediction

Predicting a next phase condition

And 5: sorting

And (4) sequencing the results obtained in the step (4), wherein the sequencing result is the sequencing situation of the event hot spots.

The method for predicting event hotspots based on internet technology as described above, wherein the step 1 includes the following steps,

calculating the network real-time attention k of a certain event a by the following formula_a

Wherein k'_aiIs the attention of a single network node i, n is all network nodes concerning event a in the entire network,

k′_aiis calculated by the following formula

Wherein k is_atInputting specific numerical values for the concerned degree of the network node at the time t from the outside;

said t₀Is the time when event a first occurs on the network;

the t is the current moment;

when t is equal to t₀The time is the time when the event a first appears on the network, and the event is an isolated time at this time, so no calculation is made.

The method for predicting event hotspots based on internet technology as described above, wherein the step 2 includes,

assuming that there are m events to predict, the probability relationship of the m events is calculated

Wherein p is calculated_jIs the probability of the jth event;

k in the formula_jIs the real-time attention of the single event calculated in step 1, and the real-time attention thus obtained for event a in step 1 is k_aWhen a specific event j is targeted, the real-time attention k is obtained_j。

The method for predicting event hotspots based on internet technology as described above, wherein the step 3 includes,

for a fixed node i, when it is concerned with an event a, k can be input_z1、k_z2、k_z3…k_zgThe g values are the attention of the node with a fixed time difference before the time t in the step 1 to the event a, and

t-zg＝zg-z(g-1)＝…＝z3-z2＝z2-z1＝Q

where Q is the time difference externally input, g is the number of data extending forward, and where Q is a minute, hour, day, week or other length of time externally input, but the Q value should be guaranteed to be in timeMinimum attention k_z1It is of interest that,

the short-term attention tendency S is calculated by the following formula_iThe method comprises the following specific steps

With time as the x-coordinate and k as the_zgPerforming quadratic curve fitting for the ordinate, and deriving the coefficient to obtain short-term attention trend S_i，

Then for event a, the total short-term attention trend S for the individual events is calculated by_a

Wherein n is the number of all nodes in the network.

The method for predicting event hotspots based on internet technology as described above, wherein the step 4 includes,

predicting the probability p of the next moment of the jth event using the following formula_j+1

Wherein p is_jIs the probability of the jth event calculated in step 2;

said S_jIs the short-term attention trend for the jth event computed in step 3.

The invention has the following remarkable effects: the method and the device only calculate aiming at the network event, do not need to carry out sample training in advance, and have low calculation complexity, so that the method and the device have good economy and effectiveness and meet the practical requirements. According to the method and the device, the influence degree of time on the event is fully considered, the influence degree of the past event development trend on the future is also fully considered, and the prediction is accurate.

Detailed Description

An event hotspot prediction method based on the Internet technology comprises the following steps:

step 1: calculating real-time attention of single event

Calculating the network real-time attention of a certain event a by the following formulak_a

Where k_aiIs the attention of a single network node i, and n is all network nodes concerning event a in the entire network.

k＇_aiIs calculated by the following formula

Wherein k is_atInputting specific numerical values for the concerned degree of the network node at the time t from the outside;

said t₀Is the time when event a first occurs on the network;

the t is the current moment;

when t is equal to t₀The time is the time when the event a first appears on the network, and the event is an isolated time at this time, so no calculation is made.

Step 2: attention to all events

Assuming that there are m events to predict, the probability relationship of the m events is calculated

Wherein p is calculated_jIs the probability of the jth event;

k in the formula_jIs the real-time attention of the single event calculated in step 1, and the real-time attention thus obtained for event a in step 1 is k_aWhen a specific event j is targeted, the real-time attention k is obtained_j。

And step 3: calculating short-term attention trends

For a fixed node i, when it is concerned with an event a, k can be input_z1、k_z2、k_z3…k_zgThe g values are the t time in step 1The node of the previous fixed time difference has a degree of attention to the event a, an

t-zg＝zg-z(g-1)＝…＝z3-z2＝z2-z1＝Q

Where Q is the time difference externally input, g is the number of forward extending data externally input, Q may be one minute, one hour, one day, one week or other time period, but the Q value should ensure a minimum focus k over time_z1It is significant. For example, if a star new movie promotion, when g takes 3 when the event first appears on the network on day 3, then Qmax can only take 1 day, so that Z1 is the value for the day first appearing on the network. Assuming that the event has occurred on the network for 4 days, g takes 3, Q takes 1 day, then k_z1Attention, k, for the first day_z2Attention, k, on the next day_z3Attention on the third day, k_atThe attention of the current time t. Similarly, Q may be taken to be 1 hour, so k_z1Attention k 3 hours before time t_z2Attention k 2 hours before time t_z3Attention 1 hour before time t

The short-term attention tendency S is calculated by the following formula_iThe method comprises the following specific steps

With time as the x-coordinate and k as the_zgPerforming quadratic curve fitting for the ordinate, and deriving the coefficient to obtain short-term attention trend S_i。

Then for event a, the total short-term attention trend S for the individual events is calculated by_a

Wherein n is the number of all nodes in the network

And 4, step 4: prediction

Predicting the probability p of the next moment of the jth event using the following formula_j+1

Whereinp_jIs the probability of the jth event calculated in step 2;

said S_jIs the short-term attention trend for the jth event computed in step 3;

and 5: sorting

For the probability p obtained in step 4_j+1And (4) sorting, wherein the sorting result is the sorting condition of the event hot spots.

Claims (5)

1. An event hotspot prediction method based on the Internet technology is characterized by comprising the following steps:

step 1: calculating real-time attention of single event

Accumulating the real-time attention degrees of all the networks at the same time;

step 2: attention to all events

Respectively calculating the attention degrees of all events needing attention in the network;

and step 3: calculating short-term attention trends

Calculating a short-term attention trend of the event;

and 4, step 4: prediction

Predicting a next phase condition

And 5: sorting

And (4) sequencing the results obtained in the step (4), wherein the sequencing result is the sequencing situation of the event hot spots.

2. The internet-technology-based event hotspot prediction method of claim 1, wherein: the step 1 includes the following steps,

calculating the network real-time attention k of a certain event a by the following formula_a

Wherein k'_aiIs the attention of a single network node i, n is all network nodes concerning event a in the entire network,

k′_aiis calculated by the following formula

Wherein k is_atInputting specific numerical values for the concerned degree of the network node at the time t from the outside;

said t₀Is the time when event a first occurs on the network;

the t is the current moment;

when t is equal to t₀The time is the time when the event a first appears on the network, and the event is an isolated time at this time, so no calculation is made.

3. The internet-technology-based event hotspot prediction method of claim 2, wherein: the step 2 comprises the steps of (a) preparing,

assuming that there are m events to predict, the probability relationship of the m events is calculated

Wherein p is calculated_jIs the probability of the jth event;

k in the formula_jIs the real-time attention of the single event calculated in step 1, and the real-time attention thus obtained for event a in step 1 is k_aWhen a specific event j is targeted, the real-time attention k is obtained_j。

4. The internet-technology-based event hotspot prediction method of claim 3, wherein: the step 3 comprises the steps of,

for a fixed node i, when it is concerned with an event a, k can be input_z1、k_z2、k_z3…k_zgThe g values are the attention of the node with a fixed time difference before the time t in the step 1 to the event a, and

t-zg＝zg-z(g-1)＝…＝z3-z2＝z2-z1＝Q

where Q is the time difference externally input, g is the number of forward extending data externally input, Q may be one minute, one hour, one day, one week or other time period, but the Q value should ensure a minimum focus k over time_z1It is of interest that,

the short-term attention tendency S is calculated by the following formula_iThe method comprises the following specific steps

With time as the x-coordinate and k as the_zgPerforming quadratic curve fitting for the ordinate, and deriving the coefficient to obtain short-term attention trend S_i，

Then for event a, the total short-term attention trend S for the individual events is calculated by_a

Wherein n is the number of all nodes in the network.

5. The internet-technology-based event hotspot prediction method of claim 4, wherein: the step 4 comprises the steps of,

predicting the probability p of the next moment of the jth event using the following formula_j+1

Wherein p is_jIs the probability of the jth event calculated in step 2;

said S_jIs the short-term attention trend for the jth event computed in step 3.