CN106355346A - Method and system for predicting box-office performance of movies on basis of neural network algorithms - Google Patents

Abstract

The invention discloses a method and a system for predicting the box-office performance of movies on the basis of neural network algorithms, and belongs to the field of technologies for processing data. The method includes steps of 1, selecting M box-office performance influence factors; 2, analyzing correlation according to box-office performance influence indexes the M box-office performance influence factors of offline movies and the box-office performance of the offline movies; 3, determining the numbers N of neurons of input layers of neural networks and the numbers of neurons of output layers; 4, determining box-office performance influence indexes of N box-office performance main influence factors of the to-be-predicted movies and determining the predicted box-office performance of the to-be-predicted movies according to output results of the neural networks. The numbers K of the neurons of the output layers are the numbers of sections of the box-office performance. The method and the system have the advantages that the integral intact system for predicting the box-office performance of the movies is established, is accurate in prediction and has positive significance on development of the movie industry in China.

Description

A kind of box office receipts Forecasting Methodology based on neural network algorithm and prognoses system

Technical field

The present invention relates to technical field of data processing, more particularly, to a kind of box office receipts prediction based on neural network algorithm Method and prognoses system.

Background technology

In recent years, Chinese film box office amplification is very high.The active support of the government and market enliven situation, show China's electricity Shadow industry size goes from strength to strength, and enters the stage of rapid growth.

The favourable of industrial environment attracts a large amount of capital to pour in cinematic industry, and the Internet capital enters film investment field comprehensively Make the investment field in industry more polynary, the investment of film channel is more abundant.

Although cinematic industry development is so rapid, what we also must regain consciousness sees the investment an industry emergence behind With the risk managed.

Meanwhile, as the important indicator of measured film quality, the effect played in cinematic industry is also to pass for box office receipts Important.The success or failure of film of the profit and loss direct reaction of box office receipts, decide income and the popularity of film.

So being predicted contributing to guaranteeing the income of film, also contribute to the operation effect of film to box office receipts simultaneously Rate.Some investment companies are provided with to gambling agreement, it helps guarantee its income.

Therefore, set up complete set and predict that accurate box office receipts prognoses system becomes China's cinematic industry development The task of top priority.

Content of the invention

In order to solve the problems, such as in prior art cannot scientific forecasting box office receipts, the invention provides a kind of based on nerve The box office receipts Forecasting Methodology of network algorithm and prognoses system.

A kind of box office receipts Forecasting Methodology based on neural network algorithm, comprising:

Step 1, selects m box office influence factor；Inquire about and calculate m box office influence factor's of multiple offline films Box office power of influence index；

Step 2, the box office power of influence index of the m box office influence factor according to described offline film and described offline The box office of film carries out correlation analysiss, obtains the correlation coefficient of m box office influence factor and box office receipts, from m box office shadow Select the larger box office influence factor of n correlation coefficient in the factor of sound as n box office major influence factors, wherein, n less than or Equal to m；

Step 3, determines that the neuron number of the input layer of neutral net is n, the neuron number of output layer is box office Segmentation number is k；The neuron number k of the neuron number n according to described input layer and described output layer determines described nerve The neuron number of the hidden layer of network is l；

The box office power of influence index of the n box office major influence factors according to the plurality of offline film trains described god Through network until forecast error is respectively less than predetermined threshold value, then train successfully；

Step 4, determines the box office power of influence index of n box office major influence factors of film to be predicted, by this n box office The box office power of influence index of major influence factors inputs to training successful neutral net as input data, according to neutral net Output result determine the prediction box office of film to be predicted.

The optimal technical scheme of box office receipts Forecasting Methodology of the present invention is to carry out neutral net instruction in described step 3 The method practiced includes:

Step 301, the number of setting input layer is n, and the number of output layer neuron is k；According to the plurality of The box office data of offline film determines the segmentation at box office, is divided into k section, and the plurality of offline film is classified, will The offline film that box office data belongs to same segmentation is divided into a class；Determine k desired output vector e, each desired output is sweared Amount comprises k element；Connection weight ω between initialization input layer and hidden layer_ij, the connection between input layer and output layer Weights ω_jk, hidden layer threshold value a, output layer threshold value b, learning rate η and neuron excitation function f (x)；

Step 302: input n box office major influence factors x of offline film, calculating hidden layer output h:

Step 303: h is exported according to hidden layer_j, connection weight ω_jkWith threshold value b, that calculate neutral net is output o:

Step 304: the output o according to neutral net and desired output e, calculating network forecast error e

e_k=e_k-o_k, k=1,2 .., k

Step 305: according to neural network forecast error e_kUpdate network connection weights ω_ij, ω_jk

ω_jk=ω_jk+ηh_je_kJ=1,2 ..., l；K=1,2 ..., k

Step 306: network node threshold value a, b are updated according to neural network forecast error e

b_k=b_k+e_k；K=1,2 ..., k

Step 307: whether evaluation algorithm iteration terminates i.e. n box office major influence factors x to all offline films Neural network forecast error whether be respectively less than pre-determined threshold, if it is, training successfully, if not, return to step 302.

The optimal technical scheme of box office receipts Forecasting Methodology of the present invention is, described box office influence factor include following because Multiple in element: film types, whether original, ip whether adapt, sequel whether, show natural law, box office of first week, public praise are scored, Secondary distribution, spin-off whether, overseas income whether, prize-winning situation, k-th director, the first leading man, the i-th leading man, first Female's protagonist, i-th female's protagonist, producer, playwright, screenwriter, manufacturing company, distributing and releasing corporation, distribution working days, web search index, relate to friendship network Searchable index, same period competition piece number, same type competition piece number, wherein k is the integer more than 0, and wherein i is the integer more than 1.

The optimal technical scheme of box office receipts Forecasting Methodology of the present invention is that the box office of described box office influence factor affects The computational methods of power index include:

When described box office factor is film types, box office power of influence index is that this film of first Zhou Zhongyu of showing of film belongs to The business of the number gained divided by the film belonging to same type with this film for the box office sum of each film of same type；

When whether original described box office factor be, when film is original, box office power of influence index is 1, when film is non-original Box office power of influence index is 0；

When whether described box office factor adapts for property right, when film is adapted for property right, power of influence index in box office is 1, and film is During non-property right reorganization, box office power of influence index is 0；

Described box office factor be sequel whether when, film is that then power of influence index in box office is 1 to sequel, film be not sequel then Box office power of influence index is 0；

When described box office factor is to show natural law, box office power of influence index shows natural law for film；

Headed by the factor of described box office during all box offices, box office power of influence index is the box office value in the first week during movie show；

When described box office factor scores for public praise, box office power of influence index is to film at least two film comment websites Scoring average；

Described box office factor be secondary distribution whether when, film is box office power of influence index when the film of secondary distribution For 1, film is not that box office power of influence index is 0 when the film of secondary distribution；

Described box office factor be spin-off whether when, it is 1 that film has power of influence index in box office during spin-off, and film does not spread out During health product, power of influence index in box office is 0；

Described box office factor be overseas income whether when, it is 1 that film has power of influence index in box office during overseas income, and film does not have Power of influence index in box office during overseas income is had to be 0；

When described box office factor is prize-winning situation, the number that box office power of influence index is won a prize for film；

When described box office factor is k-th director, box office power of influence index directs all films once led for this The meansigma methodss at first all box offices；

When described box office factor is the first leading man, box office power of influence index be this first leading man once drilled all The meansigma methodss at first all box offices of film；

When described box office factor is the i-th leading man, box office power of influence index is each electricity that this i-th leading man was once drilled The number of the film that the box office of first week of shadow was once drilled with its sum of products of coefficient of taking part in a performance in this film and this i-th leading man The business of amount；Wherein i is the integer more than 1；

When described box office factor is that the first female acts the leading role, box office power of influence index is that this first female acts the leading role once drilled all The meansigma methodss at first all box offices of film；

When described box office factor is that the i-th female acts the leading role, box office power of influence index is that this i-th female acts the leading role each electricity once drilled The number of the film that the box office of first week of shadow was once drilled with this i-th female protagonist with its sum of products of coefficient of taking part in a performance in this film The business of amount；Wherein i is the integer more than 1；

When described box office factor is producer, the first week of all films that box office power of influence index once made for producer The meansigma methodss at box office；

When described box office factor is playwright, screenwriter, box office power of influence index is the box office of first week of the once warp-knit all films of playwright, screenwriter Meansigma methodss；

When described box office factor is manufacturing company, box office power of influence index is the electricity made by all manufacturing companies of film The total number of shadow and the business of the total number of manufacturing company；

When described box office factor is distributing and releasing corporation, the electricity that box office power of influence index is issued by all distributing and releasing corporations of film The total number of shadow and the business of the total number of distributing and releasing corporation；

When described box office factor is the distribution working days, box office power of influence index is the working days classification belonging to the distribution working days of film Under all same working days classifications film first week box office with the business of the natural law of working days classification；

When described box office factor is web search index, box office power of influence index is the first search network during movie show On search week meansigma methodss；

When described box office factor is to relate to friendship web search index, box office power of influence index relates to for during movie show first Hand over the search odd-numbered day maximum of network；

Described box office factor competes piece number for the same period, and box office power of influence index is film in the same competition piece showing period Number；

Described box office factor competes piece number for same type, and box office power of influence index shows the similar of period for film same Type competes the number of piece.

The invention still further relates to a kind of box office receipts prognoses system based on neural network algorithm, comprising: the first computing module, Correlating module, selecting module, the second computing module, ANN Control module；

Described first computing module is used for selecting m box office influence factor；M inquiring about and calculating multiple offline films The box office power of influence index of box office influence factor；

Described correlating module is used for the box office power of influence of the m box office influence factor according to described offline film The box office of index and described offline film carries out correlation analysiss, obtains m box office influence factor related to box office receipts Coefficient,

Described selecting module, for selecting the larger box office influence factor of n correlation coefficient from m box office influence factor As n box office major influence factors, wherein, n is less than or equal to m；

Described second computing module, the box office shadow of the n box office major influence factors according to the plurality of offline film Ring power index and train described neutral net until forecast error is respectively less than predetermined threshold value, then train successfully；

Described ANN Control module, for determining the box office shadow of n box office major influence factors of film to be predicted Ring power index, the box office power of influence index of this n box office major influence factors is inputted to training successfully as input data Neutral net, determines the prediction box office of film to be predicted according to the output result of neutral net.

The optimal technical scheme of box office receipts prognoses system of the present invention is, described ANN Control module, according to The method that following methods carry out neural metwork training includes:

Step 301, the number of setting input layer is n, and the number of output layer neuron is k；According to the plurality of The box office data of offline film determines the segmentation at box office, is divided into k section, and the plurality of offline film is classified, will The offline film that box office data belongs to same segmentation is divided into a class；Determine k desired output vector e, each desired output is sweared Amount comprises k element；Connection weight ω between initialization input layer and hidden layer_ij, the connection between input layer and output layer Weights ω_jk, hidden layer threshold value a, output layer threshold value b, learning rate η and neuron excitation function f (x)；

Step 302: input n box office major influence factors x of offline film, calculating hidden layer output h:

Step 303: h is exported according to hidden layer_j, connection weight ω_jkWith threshold value b, that calculate neutral net is output o:

Step 304: the output o according to neutral net and desired output e, calculating network forecast error e

e_k=e_k-o_k, k=1,2 .., k

Step 305: according to neural network forecast error e_kUpdate network connection weights ω_ij, ω jk

ω_jk=ω_jk+ηh_je_kJ=1,2 ..., l；K=1,2 ..., k

Step 306: network node threshold value a, b are updated according to neural network forecast error e

b_k=b_k+e_k；K=1,2 ..., k

Step 307: whether evaluation algorithm iteration terminates i.e. n box office major influence factors x to all offline films Neural network forecast error whether be respectively less than pre-determined threshold, if it is, training successfully, if not, return to step 302.

The optimal technical scheme of box office receipts prognoses system of the present invention is, described box office influence factor include following because Multiple in element: film types, whether original, ip whether adapt, sequel whether, show natural law, box office of first week, public praise are scored, Secondary distribution, spin-off whether, overseas income whether, prize-winning situation, k-th director, the first leading man, the i-th leading man, first Female's protagonist, i-th female's protagonist, producer, playwright, screenwriter, manufacturing company, distributing and releasing corporation, distribution working days, web search index, relate to friendship network Searchable index, same period competition piece number, same type competition piece number, wherein k is the integer more than 0, and wherein i is the integer more than 1.

The optimal technical scheme of box office receipts prognoses system of the present invention is, described first computing module, is additionally operable to root The box office power of influence index of the box office influence factor of under type calculating according to this:

When described box office factor is film types, box office power of influence index is that this film of first Zhou Zhongyu of showing of film belongs to The business of the number gained divided by the film belonging to same type with this film for the box office sum of each film of same type；

When whether original described box office factor be, when film is original, box office power of influence index is 1, when film is non-original Box office power of influence index is 0；

When whether described box office factor adapts for property right, when film is adapted for property right, power of influence index in box office is 1, and film is During non-property right reorganization, box office power of influence index is 0；

Described box office factor be sequel whether when, film is that then power of influence index in box office is 1 to sequel, film be not sequel then Box office power of influence index is 0；

When described box office factor is to show natural law, box office power of influence index shows natural law for film；

Headed by the factor of described box office during all box offices, box office power of influence index is the box office value in the first week during movie show；

When described box office factor scores for public praise, box office power of influence index is to film at least two film comment websites Scoring average；

Described box office factor be secondary distribution whether when, film is box office power of influence index when the film of secondary distribution For 1, film is not that box office power of influence index is 0 when the film of secondary distribution；

Described box office factor be spin-off whether when, it is 1 that film has power of influence index in box office during spin-off, and film does not spread out During health product, power of influence index in box office is 0；

Described box office factor be overseas income whether when, it is 1 that film has power of influence index in box office during overseas income, and film does not have Power of influence index in box office during overseas income is had to be 0；

When described box office factor is prize-winning situation, the number that box office power of influence index is won a prize for film；

When described box office factor is k-th director, box office power of influence index directs all films once led for this The meansigma methodss at first all box offices；

When described box office factor is the first leading man, box office power of influence index be this first leading man once drilled all The meansigma methodss at first all box offices of film；

When described box office factor is the i-th leading man, box office power of influence index is each electricity that this i-th leading man was once drilled The number of the film that the box office of first week of shadow was once drilled with its sum of products of coefficient of taking part in a performance in this film and this i-th leading man The business of amount；Wherein i is the integer more than 1；

When described box office factor is that the first female acts the leading role, box office power of influence index is that this first female acts the leading role once drilled all The meansigma methodss at first all box offices of film；

When described box office factor is that the i-th female acts the leading role, box office power of influence index is that this i-th female acts the leading role each electricity once drilled The number of the film that the box office of first week of shadow was once drilled with this i-th female protagonist with its sum of products of coefficient of taking part in a performance in this film The business of amount；Wherein i is the integer more than 1；

When described box office factor is producer, the first week of all films that box office power of influence index once made for producer The meansigma methodss at box office；

When described box office factor is playwright, screenwriter, box office power of influence index is the box office of first week of the once warp-knit all films of playwright, screenwriter Meansigma methodss；

When described box office factor is manufacturing company, box office power of influence index is the electricity made by all manufacturing companies of film The total number of shadow and the business of the total number of manufacturing company；

When described box office factor is distributing and releasing corporation, the electricity that box office power of influence index is issued by all distributing and releasing corporations of film The total number of shadow and the business of the total number of distributing and releasing corporation；

When described box office factor is the distribution working days, box office power of influence index is the working days class belonging to the distribution working days of film First all box offices of the film of not lower all same working days classifications and the business of the natural law of working days classification；

When described box office factor is web search index, box office power of influence index is the first search network during movie show On search week meansigma methodss；

When described box office factor is to relate to friendship web search index, box office power of influence index relates to friendship for during movie show first The search odd-numbered day maximum of network；

Described box office factor competes piece number for the same period, and box office power of influence index is film in the same competition piece showing period Number；

Described box office factor competes piece number for same type, and box office power of influence index shows the similar of period for film same Type competes the number of piece.

Therefore, the box office receipts Forecasting Methodology based on neural network algorithm according to present invention offer and system, Ke Yili Accurately predict box office receipts with big data, contribute to guaranteeing the investment return of film.

Brief description

Fig. 1 is the flow chart of the box office receipts Forecasting Methodology in embodiment based on neural network algorithm；

Fig. 2 is the structure chart of the box office receipts prognoses system in embodiment based on neural network algorithm；

Fig. 3 is neural network topology structure figure.

Specific embodiment

Purpose, technical scheme and advantage for making the embodiment of the present invention are clearer, below in conjunction with the embodiment of the present invention In accompanying drawing, the technical scheme in the embodiment of the present invention is clearly and completely described it is clear that described embodiment is The a part of embodiment of the present invention, rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art The every other embodiment being obtained under the premise of not making creative work, broadly falls into the scope of protection of the invention.Need Illustrate, in the case of not conflicting, the embodiment in the application and the feature in embodiment can mutual combination in any.

Fig. 1 is the flow chart of the box office receipts Forecasting Methodology based on neural network algorithm, and the method includes:

Step 1, selects m box office influence factor；Inquire about and calculate m box office influence factor's of multiple offline films Box office power of influence index；

Step 2, the box office power of influence index of the m box office influence factor according to offline film and offline The box office of film carries out correlation analysiss, obtains the correlation coefficient of m box office influence factor and box office receipts, from m box office shadow Select the larger box office influence factor of n correlation coefficient in the factor of sound as n box office major influence factors, wherein, n less than or Equal to m；

Step 3, determines that the neuron number of the input layer of neutral net is n, the neuron number of output layer is box office Segmentation number is k；The neuron number k of the neuron number n according to input layer and output layer determines the hidden layer of neutral net Neuron number be l；The box office power of influence index instruction of the n box office major influence factors according to multiple offline films Practice neutral net until forecast error is respectively less than predetermined threshold value, then train successfully；

Step 4, determines the box office power of influence index of n box office major influence factors of film to be predicted, by this n box office The box office power of influence index of major influence factors inputs to training successful neutral net as input data, according to neutral net Output result determine the prediction box office of film to be predicted.

Wherein, the method carrying out neural metwork training in step 3 includes:

Step 301, the number of setting input layer is n, and the number of output layer neuron is k；According to multiple under The box office data of the offline film of line determines the segmentation at box office, is divided into k section, and multiple offline films are classified, will The offline film that box office data belongs to same segmentation is divided into a class；Determine k desired output vector e, each expectation is defeated Go out vector and comprise k element；Connection weight ω between initialization input layer and hidden layer_ij, between input layer and output layer Connection weight ω_jk, hidden layer threshold value a, output layer threshold value b, learning rate η and neuron excitation function f (x)；

Step 302: input n box office major influence factors x of offline film, calculating hidden layer output h:

Step 303: h is exported according to hidden layer_j, connection weight ω_jkWith threshold value b, that calculate neutral net is output o:

Step 304: the output o according to neutral net and desired output e, calculating network forecast error e

e_k=e_k-o_k, k=1,2 .., k

Step 305: according to neural network forecast error e_kUpdate network connection weights ω_ij, ω_jk

ω_jk=ω_jk+ηh_je_kJ=1,2 ..., l；K=1,2 ..., k

Step 306: network node threshold value a, b are updated according to neural network forecast error e

b_k=b_k+e_k；K=1,2 ..., k

Step 307: whether evaluation algorithm iteration terminates i.e. n box office major influence factors x to all offline films Neural network forecast error whether be respectively less than pre-determined threshold, if it is, training successfully, if not, return to step 302.

In this method, box office influence factor includes multiple in following factors: film types, whether original, ip reorganization with No, sequel whether, show natural law, box office of first week, public praise scoring, secondary distribution, spin-off whether, overseas income whether, prize-winning Situation, k-th director, the first leading man, the i-th leading man, first female's protagonist, i-th female's protagonist, producer, playwright, screenwriter, making are public Department, distributing and releasing corporation, distribution working days, web search index, relate to friendship web search index, the same period competition piece number, same type competition piece Number, wherein k is the integer more than 0, and wherein i is the integer more than 1.

Determine that the box office power of influence of box office influence factor refers to calibration method and specifically includes:

When box office factor is film types, box office power of influence index is that this film of first Zhou Zhongyu of showing of film belongs to identical The business of the number gained divided by the film belonging to same type with this film for the box office sum of each film of type；Specifically, ticket When room factor is t-th film types, box office power of influence index is y_t

Wherein t represents the sequence number of film types, and m1 represents the film total amount of t type, and j represents the electricity belonging to t type The sequence number of shadow, a_tjRepresent the box office that the jth portion film of content t type produced within the first week shown.

When whether original box office factor be, when film is original, box office power of influence index is 1, box office when film is non-original Power of influence index is 0；

When whether box office factor adapts for property right, when film is adapted for property right, power of influence index in box office is 1, and film is non-product During power reorganization, box office power of influence index is 0；

Box office factor be sequel whether when, film is that then power of influence index in box office is 1 to sequel, and film is not sequel then box office Power of influence index is 0；

When box office factor is to show natural law, box office power of influence index shows natural law for film；

Headed by the factor of box office during all box offices, box office power of influence index is the box office value in the first week during movie show；

When box office factor scores for public praise, box office power of influence index is at least two film comment websites, film to be commented The average divided；

Box office factor be secondary distribution whether when, film is that box office power of influence index is 1 when the film of secondary distribution, Film is not that box office power of influence index is 0 when the film of secondary distribution；

Box office factor be spin-off whether when, it is 1 that film has power of influence index in box office during spin-off, and film does not have spin-off When box office power of influence index be 0；

Box office factor be overseas income whether when, it is 1 that film has power of influence index in box office during overseas income, and film does not have During overseas income, box office power of influence index is 0；

When box office factor is prize-winning situation, the number that box office power of influence index is won a prize for film；

When box office factor is k-th director, box office power of influence index directs the first week of all films once led for this The meansigma methodss at box office；Wherein k is the integer more than 0；

Specifically box office power of influence index is

Wherein k represents the sequence number of director, and m2 represents the film total amount of k-th director's shooting, and j represents that k-th director participates in The sequence number of the film shooting, a_kjRepresent that k-th director participates in the box office that the jth portion film of shooting produced within the first week shown.

When box office factor is the first leading man, box office power of influence index is all films that this first leading man was once drilled First week box office meansigma methodss；

When box office factor is the i-th leading man, box office power of influence index is each film that this i-th leading man was once drilled The quantity of film that first all box offices were once drilled with its sum of products of coefficient of taking part in a performance in this film and this i-th leading man Business；Wherein i is the integer more than 1；

Specifically box office power of influence index is

Wherein i represents the sequence number of leading man, and m3 represents the film total amount that the i-th leading man shoots, and j represents the i-th leading man ginseng With the sequence number of the film shooting, a_ijRepresent the ticket that the jth portion film that the i-th leading man participates in shooting produced within the first week shown Room, u_ijRepresent coefficient of taking part in a performance in jth portion film for i-th leading man, u_ij=1- (n_ij- 1)/5, n_ijRepresent the i-th leading man Protagonist order in j portion film.Coefficient of taking part in a performance in current movie for 2nd leading man of such as current movie is 0.8.

When box office factor is that the first female acts the leading role, box office power of influence index is all films that this first female protagonist was once drilled First week box office meansigma methodss；

When box office factor is that the i-th female acts the leading role, box office power of influence index is each film that this i-th female protagonist was once drilled The quantity of film that first all box offices were once drilled with this i-th female protagonist with its sum of products of coefficient of taking part in a performance in this film Business；Wherein i is the integer more than 1.The calculation of box office power of influence index of corresponding box office factor and the calculating of the i-th leading man Mode is identical, and here is omitted.

When box office factor is producer, the box office of first week of all films that box office power of influence index once made for producer Meansigma methodss；

When box office factor is playwright, screenwriter, box office power of influence index is the average of first all box offices of the once warp-knit all films of playwright, screenwriter Value；

When box office factor is manufacturing company, box office power of influence index is the film made by all manufacturing companies of film The business of the total number of total number and manufacturing company；

When box office factor is distributing and releasing corporation, the film that box office power of influence index is issued by all distributing and releasing corporations of film The business of the total number of total number and distributing and releasing corporation；

When box office factor is the distribution working days, institute under working days classification belonging to the distribution working days of film for the box office power of influence index There are first all box offices of film of same working days classification and the business of the natural law of working days classification；

When box office factor is web search index, box office power of influence index is on the first search network during movie show The all meansigma methodss of search；

When box office factor is to relate to friendship web search index, box office power of influence index relates to friendship network for during movie show first Search odd-numbered day maximum；

Box office factor competes piece number for the same period, and box office power of influence index is that film is individual in the same competition piece showing period Number；

Box office factor is same type competition piece number, and box office power of influence index is film to show the same type in period competing same Strive the number of piece.

Fig. 3 is neural network topology structure figure, wherein, x₁,x₂...x_nIt is the input value of bp neutral net, y₁,y₂...y_k It is the predictive value of bp neutral net, ω_ij, ω_jkWeights for bp neutral net.

Fig. 2 is the structure chart of the box office receipts prognoses system based on neural network algorithm, and this system includes: the first calculating mould Block, correlating module, selecting module, the second computing module, ANN Control module.

First computing module is used for selecting m box office influence factor；Inquire about and calculate m ticket of multiple offline films The box office power of influence index of room influence factor；

Correlating module is used for the box office power of influence index of m box office influence factor according to offline film and The box office of offline film carries out correlation analysiss, obtains the correlation coefficient of m box office influence factor and box office receipts,

Selecting module, for selecting the larger box office influence factor's conduct of n correlation coefficient from m box office influence factor N box office major influence factors, wherein, n is less than or equal to m；

Second computing module, the box office power of influence index of the n box office major influence factors according to multiple offline films Training neutral net until forecast error is respectively less than predetermined threshold value, is then trained successfully；

ANN Control module, for determining the box office power of influence of n box office major influence factors of film to be predicted Index, the box office power of influence index of this n box office major influence factors is inputted to training successfully nerve as input data Network, determines the prediction box office of film to be predicted according to the output result of neutral net.

Wherein,

ANN Control module includes according to the method that following methods carry out neural metwork training:

Step 301, the number of setting input layer is n, and the number of output layer neuron is k；According to multiple under The box office data of line film determines the segmentation at box office, is divided into k section, multiple offline films is classified, by box office data The offline film belonging to same segmentation is divided into a class；Determine k desired output vector e, each desired output vector comprises k Element；Connection weight ω between initialization input layer and hidden layer_ij, connection weight ω between input layer and output layer_jk、 Hidden layer threshold value a, output layer threshold value b, learning rate η and neuron excitation function f (x)；

Step 302: input n box office major influence factors x of offline film, calculating hidden layer output h:

Step 303: h is exported according to hidden layer_j, connection weight ω_jkWith threshold value b, that calculate neutral net is output o:

Step 304: the output o according to neutral net and desired output e, calculating network forecast error e

e_k=e_k-o_k, k=1,2 .., k

Step 305: according to neural network forecast error e_kUpdate network connection weights ω_ij, ω_jk

ω_jk=ω_jk+ηh_je_kJ=1,2 ..., l；K=1,2 ..., m

Step 306: network node threshold value a, b are updated according to neural network forecast error e

b_k=b_k+e_k；K=1,2 ..., k

Step 307: whether evaluation algorithm iteration terminates i.e. n box office major influence factors x to all offline films Neural network forecast error whether be respectively less than pre-determined threshold, if it is, training successfully, if not, return to step 302.

Box office influence factor includes multiple in following factors: film types, whether original, whether ip adapts, sequel with No, show natural law, box office of first week, public praise scoring, secondary distribution, spin-off whether, overseas income whether, prize-winning situation, k-th Director, the first leading man, the i-th leading man, first female's protagonist, i-th female's protagonist, producer, playwright, screenwriter, manufacturing company, distributing and releasing corporation, Issue working days, web search index, relate to friendship web search index, same period competition piece number, same type competition piece number, wherein k is big In 0 integer, wherein i is the integer more than 1.

First computing module calculates described in the mode of box office power of influence index and the said method of box office influence factor Identical, here is omitted.

The present invention can accurately predict box office receipts using big data.

Descriptions above can combine individually or in every way enforcement, and these variant are all Within protection scope of the present invention.

Herein, term " inclusion ", "comprising" or its any other variant are intended to comprising of nonexcludability, from And make to include a series of article of key elements or equipment not only includes those key elements, but also inclusion be not expressly set out its His key element, or also include for this article or the intrinsic key element of equipment.In the absence of more restrictions, by language Key element that sentence " include ... " limits it is not excluded that also exist in the article including key element or equipment other identical will Element.

Above example only in order to technical scheme to be described and unrestricted, reference only to preferred embodiment to this Bright it has been described in detail.It will be understood by those within the art that, technical scheme can be modified Or equivalent, without deviating from the spirit and scope of technical solution of the present invention, all should cover the claim model in the present invention In the middle of enclosing.

Claims (8)

1. a kind of box office receipts Forecasting Methodology based on neural network algorithm is it is characterised in that include:

Step 1, selects m box office influence factor；Inquire about and calculate the box office of m box office influence factor of multiple offline films Power of influence index；

Step 2, the box office power of influence index of the m box office influence factor according to described offline film and described offline film Box office carry out correlation analysiss, obtain the correlation coefficient of m box office influence factor and box office receipts, from the impact of m box office because Select the larger box office influence factor of n correlation coefficient as n box office major influence factors in element, wherein, n is less than or equal to m；

Step 3, determines that the neuron number of the input layer of neutral net is n, the neuron number of output layer is the segmentation at box office Number is k；The neuron number k of the neuron number n according to described input layer and described output layer determines described neutral net Hidden layer neuron number be l；

The box office power of influence index of the n box office major influence factors according to the plurality of offline film trains described nerve net Network until forecast error is respectively less than predetermined threshold value, is then trained successfully；

Step 4, determines the box office power of influence index of n box office major influence factors of film to be predicted, will be main for this n box office The box office power of influence index of influence factor inputs to training successful neutral net as input data, defeated according to neutral net Go out the prediction box office that result determines film to be predicted.

2. the box office receipts Forecasting Methodology based on neural network algorithm as claimed in claim 1 it is characterised in that

The method carrying out neural metwork training in described step 3 includes:

Step 301, the number of setting input layer is n, and the number of output layer neuron is k；According to the plurality of under The box office data of line film determines the segmentation at box office, is divided into k section, the plurality of offline film is classified, by box office The offline film that data belongs to same segmentation is divided into a class；Determine k desired output vector e, each desired output vector bag Containing k element；Connection weight ω between initialization input layer and hidden layer_ij, the connection weight between input layer and output layer ω_jk, hidden layer threshold value a, output layer threshold value b, learning rate η and neuron excitation function f (x)；

Step 302: input n box office major influence factors x of offline film, calculating hidden layer output h:

$h_j=f(\underset{i=1}{\overset{n}{\mathrm{\σ}}}{\mathrm{\ω}}_{ij}{x}_i-a_j),i=1,2,...,n;j=1,2,...,l$

Step 303: h is exported according to hidden layer_j, connection weight ω_jkWith threshold value b, that calculate neutral net is output o:

$o_k=\underset{j=1}{\overset{l}{\mathrm{\σ}}}{h}_j{\mathrm{\ω}}_{jk}-b_k,j=1,2,...,l;k=1,2,...,k$

Step 304: the output o according to neutral net and desired output e, calculating network forecast error e

e_k=e_k-o_k, k=1,2 .., k

Step 305: according to neural network forecast error e_kUpdate network connection weights ω_ij, ω_jk

${\mathrm{\ω}}_{ij}={\mathrm{\ω}}_{ij}+{\mathrm{\ηh}}_j(1-h_j)x\left(i\right)\underset{k=1}{\overset{k}{\σ}}{\mathrm{\ω}}_{ij}{e}_k,i=1,2,...,n;j=1,2,...,l;k=1,2,...,k$

ω_jk=ω_jk+ηh_je_kJ=1,2 ..., l；K=1,2 ..., k

Step 306: network node threshold value a, b are updated according to neural network forecast error e

$a_j=a_j+{\mathrm{\ηh}}_j(1-h_j)\underset{k=1}{\overset{k}{\σ}}{\mathrm{\ω}}_{jk}{e}_k;j=1,2,...,l;k=1,2,...,k$

b_k=b_k+e_k；K=1,2 ..., k

Step 307: whether evaluation algorithm iteration terminates the net of i.e. n box office major influence factors x to all offline films Whether network forecast error is respectively less than pre-determined threshold, if it is, training successfully, if not, return to step 302.

3. the box office receipts Forecasting Methodology based on neural network algorithm as claimed in claim 1 it is characterised in that

Described box office influence factor includes multiple in following factors: film types, whether original, whether ip adapts, sequel with No, show natural law, box office of first week, public praise scoring, secondary distribution, spin-off whether, overseas income whether, prize-winning situation, k-th Director, the first leading man, the i-th leading man, first female's protagonist, i-th female's protagonist, producer, playwright, screenwriter, manufacturing company, distributing and releasing corporation, Issue working days, web search index, relate to friendship web search index, same period competition piece number, same type competition piece number, wherein k is big In 0 integer, wherein i is the integer more than 1.

4. the box office receipts Forecasting Methodology based on neural network algorithm as claimed in claim 1 or 2 it is characterised in that

The computational methods of the box office power of influence index of described box office influence factor include:

When described box office factor is film types, box office power of influence index is that this film of first Zhou Zhongyu of showing of film belongs to identical The business of the number gained divided by the film belonging to same type with this film for the box office sum of each film of type；

When whether original described box office factor be, when film is original, box office power of influence index is 1, box office when film is non-original Power of influence index is 0；

When whether described box office factor adapts for property right, when film is adapted for property right, power of influence index in box office is 1, and film is non-product During power reorganization, box office power of influence index is 0；

Described box office factor be sequel whether when, film is that then power of influence index in box office is 1 to sequel, and film is not sequel then box office Power of influence index is 0；

When described box office factor is to show natural law, box office power of influence index shows natural law for film；

Headed by the factor of described box office during all box offices, box office power of influence index is the box office value in the first week during movie show；

When described box office factor scores for public praise, box office power of influence index is at least two film comment websites, film to be commented The average divided；

Described box office factor be secondary distribution whether when, film is that box office power of influence index is 1 when the film of secondary distribution, Film is not that box office power of influence index is 0 when the film of secondary distribution；

Described box office factor be spin-off whether when, it is 1 that film has power of influence index in box office during spin-off, and film does not have spin-off When box office power of influence index be 0；

Described box office factor be overseas income whether when, it is 1 that film has power of influence index in box office during overseas income, and film does not have sea During outer income, power of influence index in box office is 0；

When described box office factor is prize-winning situation, the number that box office power of influence index is won a prize for film；

When described box office factor is k-th director, box office power of influence index directs the first week of all films once led for this The meansigma methodss at box office；

When described box office factor is the first leading man, box office power of influence index is all films that this first leading man was once drilled First week box office meansigma methodss；

When described box office factor is the i-th leading man, box office power of influence index is each film that this i-th leading man was once drilled The quantity of film that first all box offices were once drilled with its sum of products of coefficient of taking part in a performance in this film and this i-th leading man Business；Wherein i is the integer more than 1；

When described box office factor is that the first female acts the leading role, box office power of influence index is all films that this first female protagonist was once drilled First week box office meansigma methodss；

When described box office factor is that the i-th female acts the leading role, box office power of influence index is each film that this i-th female protagonist was once drilled The quantity of film that first all box offices were once drilled with this i-th female protagonist with its sum of products of coefficient of taking part in a performance in this film Business；Wherein i is the integer more than 1；

When described box office factor is producer, the box office of first week of all films that box office power of influence index once made for producer Meansigma methodss；

When described box office factor is playwright, screenwriter, box office power of influence index is the average of first all box offices of the once warp-knit all films of playwright, screenwriter Value；

When described box office factor is manufacturing company, box office power of influence index is the film made by all manufacturing companies of film The business of the total number of total number and manufacturing company；

When described box office factor is distributing and releasing corporation, the film that box office power of influence index is issued by all distributing and releasing corporations of film The business of the total number of total number and distributing and releasing corporation；

When described box office factor is the distribution working days, institute under working days classification belonging to the distribution working days of film for the box office power of influence index There are first all box offices of film of same working days classification and the business of the natural law of working days classification；

When described box office factor is web search index, box office power of influence index is on the first search network during movie show The all meansigma methodss of search；

When described box office factor is to relate to friendship web search index, box office power of influence index relates to friendship network for during movie show first Search odd-numbered day maximum；

Described box office factor competes piece number for the same period, and box office power of influence index is that film is individual in the same competition piece showing period Number；

Described box office factor is same type competition piece number, and box office power of influence index is film to show the same type in period competing same Strive the number of piece.

5. a kind of box office receipts prognoses system based on neural network algorithm is it is characterised in that include: the first computing module, phase Closing property analysis module, selecting module, the second computing module, ANN Control module；

Described first computing module is used for selecting m box office influence factor；Inquire about and calculate m box office of multiple offline films The box office power of influence index of influence factor；

Described correlating module is used for the box office power of influence index of the m box office influence factor according to described offline film Carry out correlation analysiss with the box office of described offline film, obtain the correlation coefficient of m box office influence factor and box office receipts,

Described selecting module, for selecting the larger box office influence factor's conduct of n correlation coefficient from m box office influence factor N box office major influence factors, wherein, n is less than or equal to m；

Described second computing module, the box office power of influence of the n box office major influence factors according to the plurality of offline film Index trains described neutral net until forecast error is respectively less than predetermined threshold value, then train successfully；

Described ANN Control module, for determining the box office power of influence of n box office major influence factors of film to be predicted Index, the box office power of influence index of this n box office major influence factors is inputted to training successfully nerve as input data Network, determines the prediction box office of film to be predicted according to the output result of neutral net.

6. the box office receipts prognoses system based on neural network algorithm as claimed in claim 5 it is characterised in that

Described ANN Control module, includes according to the method that following methods carry out neural metwork training:

Step 301, the number of setting input layer is n, and the number of output layer neuron is k；According to the plurality of under The box office data of line film determines the segmentation at box office, is divided into k section, the plurality of offline film is classified, by box office The offline film that data belongs to same segmentation is divided into a class；Determine k desired output vector e, each desired output vector bag Containing k element；Connection weight ω between initialization input layer and hidden layer_ij, the connection weight between input layer and output layer ω_jk, hidden layer threshold value a, output layer threshold value b, learning rate η and neuron excitation function f (x)；

Step 302: input n box office major influence factors x of offline film, calculating hidden layer output h:

$h_j=f(\underset{i=1}{\overset{n}{\mathrm{\σ}}}{\mathrm{\ω}}_{ij}{x}_i-a_j),i=1,2,...,n;j=1,2,...,l$

Step 303: h is exported according to hidden layer_j, connection weight ω_jkWith threshold value b, that calculate neutral net is output o:

$o_k=\underset{j=1}{\overset{l}{\mathrm{\σ}}}{h}_j{\mathrm{\ω}}_{jk}-b_k,j=1,2,...,l;k=1,2,...,k$

Step 304: the output o according to neutral net and desired output e, calculating network forecast error e

e_k=e_k-o_k, k=1,2 .., k

Step 305: according to neural network forecast error e_kUpdate network connection weights ω_ij, ω_jk

${\mathrm{\ω}}_{ij}={\mathrm{\ω}}_{ij}+{\mathrm{\ηh}}_j(1-h_j)x_i\underset{k=1}{\overset{k}{\σ}}{\mathrm{\ω}}_{ij}{e}_k,i=1,2,...,n;j=1,2,...,l;k=1,2,...,k$

ω_jk=ω_jk+ηh_je_kJ=1,2 ..., l；K=1,2 ..., k

Step 306: network node threshold value a, b are updated according to neural network forecast error e

$a_j=a_j+{\mathrm{\ηh}}_j(1-h_j)\underset{k=1}{\overset{k}{\σ}}{\mathrm{\ω}}_{jk}{e}_k;j=1,2,...,l;k=1,2,...,k$

b_k=b_k+e_k；K=1,2 ..., k

Step 307: whether evaluation algorithm iteration terminates the net of i.e. n box office major influence factors x to all offline films Whether network forecast error is respectively less than pre-determined threshold, if it is, training successfully, if not, return to step 302.

7. the box office receipts prognoses system based on neural network algorithm as claimed in claim 5 it is characterised in that

Described box office influence factor includes multiple in following factors: film types, whether original, whether ip adapts, sequel with No, show natural law, box office of first week, public praise scoring, secondary distribution, spin-off whether, overseas income whether, prize-winning situation, k-th Director, the first leading man, the i-th leading man, first female's protagonist, i-th female's protagonist, producer, playwright, screenwriter, manufacturing company, distributing and releasing corporation, Issue working days, web search index, relate to friendship web search index, same period competition piece number, same type competition piece number, wherein k is big In 0 integer, wherein i is the integer more than 1.

8. the box office receipts prognoses system based on neural network algorithm as described in claim 5 or 6 it is characterised in that

Described first computing module, is additionally operable to calculate the box office power of influence index of box office influence factor according in the following manner:

When described box office factor is film types, box office power of influence index is that this film of first Zhou Zhongyu of showing of film belongs to identical The business of the number gained divided by the film belonging to same type with this film for the box office sum of each film of type；

When whether original described box office factor be, when film is original, box office power of influence index is 1, box office when film is non-original Power of influence index is 0；

When whether described box office factor adapts for property right, when film is adapted for property right, power of influence index in box office is 1, and film is non-product During power reorganization, box office power of influence index is 0；

Described box office factor be sequel whether when, film is that then power of influence index in box office is 1 to sequel, and film is not sequel then box office Power of influence index is 0；

When described box office factor is to show natural law, box office power of influence index shows natural law for film；

Headed by the factor of described box office during all box offices, box office power of influence index is the box office value in the first week during movie show；

When described box office factor scores for public praise, box office power of influence index is at least two film comment websites, film to be commented The average divided；

Described box office factor be secondary distribution whether when, film is that box office power of influence index is 1 when the film of secondary distribution, Film is not that box office power of influence index is 0 when the film of secondary distribution；

Described box office factor be spin-off whether when, it is 1 that film has power of influence index in box office during spin-off, and film does not have spin-off When box office power of influence index be 0；

Described box office factor be overseas income whether when, it is 1 that film has power of influence index in box office during overseas income, and film does not have sea During outer income, power of influence index in box office is 0；

When described box office factor is prize-winning situation, the number that box office power of influence index is won a prize for film；

When described box office factor is k-th director, box office power of influence index directs the first week of all films once led for this The meansigma methodss at box office；

When described box office factor is the first leading man, box office power of influence index is all films that this first leading man was once drilled First week box office meansigma methodss；

When described box office factor is the i-th leading man, box office power of influence index is each film that this i-th leading man was once drilled The quantity of film that first all box offices were once drilled with its sum of products of coefficient of taking part in a performance in this film and this i-th leading man Business；Wherein i is the integer more than 1；

When described box office factor is that the first female acts the leading role, box office power of influence index is all films that this first female protagonist was once drilled First week box office meansigma methodss；

When described box office factor is that the i-th female acts the leading role, box office power of influence index is each film that this i-th female protagonist was once drilled The quantity of film that first all box offices were once drilled with this i-th female protagonist with its sum of products of coefficient of taking part in a performance in this film Business；Wherein i is the integer more than 1；

When described box office factor is producer, the box office of first week of all films that box office power of influence index once made for producer Meansigma methodss；

When described box office factor is playwright, screenwriter, box office power of influence index is the average of first all box offices of the once warp-knit all films of playwright, screenwriter Value；

When described box office factor is manufacturing company, box office power of influence index is the film made by all manufacturing companies of film The business of the total number of total number and manufacturing company；

When described box office factor is distributing and releasing corporation, the film that box office power of influence index is issued by all distributing and releasing corporations of film The business of the total number of total number and distributing and releasing corporation；

When described box office factor is the distribution working days, institute under working days classification belonging to the distribution working days of film for the box office power of influence index There are first all box offices of film of same working days classification and the business of the natural law of working days classification；

When described box office factor is web search index, box office power of influence index is on the first search network during movie show The all meansigma methodss of search；

When described box office factor is to relate to friendship web search index, box office power of influence index relates to friendship network for during movie show first Search odd-numbered day maximum；

Described box office factor competes piece number for the same period, and box office power of influence index is that film is individual in the same competition piece showing period Number；

Described box office factor is same type competition piece number, and box office power of influence index is film to show the same type in period competing same Strive the number of piece.