CN106408217A - Product life cycle identification method and device - Google Patents

Abstract

The invention provides a product life cycle identification method and device, which can identify life cycle and improve identification accuracy. The method comprises the following steps: obtaining sales data of products in a sales system, and extracting a life cycle pattern set from the data according to a product life cycle pattern extracting method, a product life cycle stage division method and sales curve similarity calculation method; determining the position of life stage of each life cycle pattern in the life cycle pattern set according to the product life cycle stage division method; calculating similarity between a sales curve of a new product and each life cycle pattern according to the sales curve similarity calculation method to obtain a similarity set; serving the life cycle pattern corresponding to the largest similarity in the similarity set as the life cycle pattern of the new product, and recording relative movement position of the sales curve of the new product when the similarity is the largest and the life cycle pattern of the new product; and determining the life cycle of the new product according to the position of life stage of the life cycle pattern of the new product and the relative movement position.

Description

A kind of recognition methods of product life cycle and device

Technical field

The present invention relates to computer and its software technology field, particularly to a kind of recognition methods of product life cycle and Device.

Background technology

Each product has a life cycle, and the life cycle of product is divided into four parts：Introduction period, growth stage, one-tenth Ripe phase and decline phase.Decline phase can be divided into two parts further：Decline phase and end period.Product life cycle table Show the whole evolution process that product is withdrawn from the market from coming into the market to.When product is in different phase, enterprise is from marketing etc. Many aspects formulate different targets and strategy, thus realize company interest maximizing, the therefore division of product life cycle, know With prediction and its related management, suitably strategy is not formulated to an enterprise and play vital status, enterprise is only to product Life cycle have a control accurately and timely, can ensure that enterprise is in invincible position in fierce market competition.

New product from initially enter market to by the whole process of market, be divided into the introduction period, the growth stage, the maturity period, Decline phase four-stage：

Introduction period refers generally to the stage that trial sale is successfully put in trial production of new products on market, and it is mainly characterized by：1. product just enters Enter test marketing, not yet accepted by client, its sales volume slowly increases；2. production lot very little, ducts starting cost is very big, thus Production is relatively costly；3. user does not know about to product and is unfamiliar with, and needs many advertisements, and selling charges are higher；4. remove imitative Outside product, product commercially typically no intratype competition；5. product has just enter into market, due to production cost and selling charges relatively Height, enterprise is being often financially loss.

After growth stage refers to that the trial sale of new products is successful, proceed to the stage producing and expanding market sale by batch.Its master Want feature：1. sales volume promptly increases；2. product design and technique finalize the design almost, and can organize and produce in batch or in a large number, produce Product cost is remarkably decreased；3. user to product oneself through being familiar with, advertising expenditure can be relatively reduced, selling cost significantly under Fall；4. increasing rapidly with yield and sales volume, enterprise turns loss into gain, and profit rises rapidly；5. intratype competition person starts to imitate Make this kind of product, market starts Competitive Trend.

Maturity period refers to that product enters and produces in enormous quantities, and is commercially in the most hotly competitive stage.It is mainly special Levying is：Market demand has gradually tended to saturation, and sales volume peaks；2. production lot is big, and product cost is low, profit Also it is up to a lot of like product of peak 3. and has been enter into market, market competition is very fierce；4. the later stage of maturation, the market demand Reach saturation, sales growth rate levels off to zero, or even negative.

Decline phase refers to that product is gradually aging, proceeds to the New Times of model change.It is mainly characterized by：1. existing New product comes into the market, and is gradually replacing old product；2. in addition to minority or indivedual best brand of product, the market sales volume increasingly under Fall；3. what market competition projected shows as price competition, and price is constantly forced to decline.

In prior art, the recognition methods of product growth cycle is mainly had：Sales growth rate determining method, analogy judge Method, Gong Juzhong method and bass diffusion models；In the recognition methods of existing product growth cycle, sales growth rate is sentenced Disconnected method and the excessive dependence Heuristics of analogy determining method go to judge the life cycle of a product, its subjective factor is excessive, knot Fruit is not sufficiently stable.Gong Juzhong method and bass diffusion models have all formulated corresponding mathematical modulo to the life cycle of a product Type, two methods are higher for the fitting effect of the more full product of life cycle, but for being in the introduction period (for example：New product), Or sales volume fluctuation is larger (for example：Sales volume that massive promotional campaign brings increases sharply) product, because data volume is less, data time Span is shorter and data fluctuations are big etc., and reason leads to data fitting effect poor.

Content of the invention

In view of this, the present invention provides a kind of recognition methods of product life cycle and device, can be according to given The mathematics division methods of life-cycle stages divide to life-cycle stages, simultaneously the current institute to product The life cycle phase at place is identified being predicted with future time instance life cycle, not only avoid the impact of subjective factor, And new product is carried out reevaluating matching without using model to it, amount of calculation is relatively small, finally improves product The accuracy rate of product life cycle identification and efficiency.

For achieving the above object, according to an aspect of the invention, it is provided a kind of recognition methods of product life cycle.

The recognition methods of the product life cycle of the present invention includes：Obtain the sale number specifying class product in marketing system According to then that this sales data is each according to the extracting method of default product life cycle pattern, default product life cycle Between the division methods in stage and default sales volume curve, the computational methods of similarity extract the life of described specified class product Cyclic pattern set；Determine described specified class product life cycle mould according to the division methods of described Various Phases of Their Life Cycle The interval position of the life stage of each lifetime value in formula set；Produce under described specified class product is new Product, calculate described new product sales volume curve and described life cycle mould according to similarity calculating method between default sales volume curve The similarity of each lifetime value in formula set, to obtain similarity set；By similarity in described similarity set When maximum, corresponding lifetime value is as the lifetime value of described new product, and records similarity maximum when institute State the displaced posi of new product sales volume curve and described new product lifetime value；According to described new product life cycle The interval position of the life stage of pattern and described displaced posi determine the life cycle residing for this new product.

Alternatively, the described step extracting the lifetime value set specifying class product includes：

Assume to have U product in certain colony, using bass diffusion models, the sales volume curve of U product is fitted Obtain collection of curves B={ B₁,B₂,...,B_U}；

The song of B in the collection of curves that division methods according to default Various Phases of Their Life Cycle obtain to described matching Line carries out the division in life cycle each stage, so that it is determined that having the product set G of complete lifecycle in U product_full= {G₁,G₂,...,G_V, wherein V≤U；

Computational methods according to similarity between default sales volume curve calculate in V product product sales volume curve two-by-two Similarity, obtains the similarity matrix between a V product, and records corresponding two of each similarity in this similarity matrix Product sales volume curve displaced posi：

One threshold epsilon of setting, is clustered to V product using this threshold value and is closed C to obtain final gathering_final；

For any one cluster C_i∈C_final, the product collection included in cluster is combined into G_i={ G₁,G₂,…,G_I, according to this In cluster, any two product is calculating the two product sales volume curve displaced posi retaining during final similarity to I product Primary products sales volume curve be overlapped merging, the product sales volume curve after being merged is：

Wherein, O_tIt is to be t in the time, the product of product sales volume curves overlapped combines, | | O_t| | represent this set unit The number of element；

For cluster C_i, bass diffusion models alignment is reused to the product sales volume curve after merging and is fitted, obtain Cluster C_iFinal product life cycle pattern P_i=(m_i,p_i,q_i)；

Belong to C to all_finalCluster all carry out as cluster C_iOperation, thus obtain specify class product product life cycle Set of modes P={ P₁,P₂,...,P_W}.

Alternatively, the curve set that matching described in the described division methods according to default Various Phases of Their Life Cycle obtains In conjunction, the curve of B carries out the division in life cycle each stage, so that it is determined that having the product collection of complete lifecycle in U product The step closed includes steps A and step B：

Step A：The life cycle of product is corresponding by the division methods according to default Various Phases of Their Life Cycle Life stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；Step A concrete Step includes：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By maximum selling value institute on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Corresponding time point as the end point of introduction period, simultaneously as the starting point of growth stage,

By on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to maximum selling value as the end point of growth stage, simultaneously as the starting point in maturity period,

By on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to minimum sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By maximum selling value institute on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Corresponding time point as the end point of decline phase, simultaneously as the starting point of end period；

Step B：For the arbitrary product in U product it is assumed that its product original sales volume curve y₁={ x₁,x₂,…,x_n, Using Bass model matching curve B out₁={ b₁,b₂,…,b_n, by the curve after this matching according to the said goods Life Cycle The division methods in phase in each stage determine each stage of the life cycle of product, if matching curve out contains described product All stages of product life cycle are it is determined that this product has complete life cycle.

Alternatively, the described computational methods according to similarity between default sales volume curve calculate in V product and produce two-by-two The step of the similarity of product sales volume curve includes：

Assume that in V product, the time series corresponding to any two product sales volume curves is respectively：WithWherein, n >=m, table respectively Show sequences y⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾The sales volume value of the product when each moment；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value, and in the sequence of structure, front-end and back-end all include m-1 NA value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with Time series y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time Sequences y^baseI-th point of alignment when, definition similarity now is：

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus Obtain the similarity between two sales volume curves corresponding to above-mentioned two product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}

For remaining product in V product, repeat the above steps, thus it is bent to obtain in V product product sales volume two-by-two The similarity of line.

Alternatively, described V product cluster using this threshold value closes C to obtain final gathering_finalStep bag Include：

For the product complete or collected works G={ G with complete lifecycle₁,G₂,…G_v}：

Step a：A product G is chosen from described product complete or collected works G_i, as first cluster C₁={ G_iIn product；

Step b：By G_iRemove from product complete or collected works, take out and in product complete or collected works G, remove G_iSurplus products G_others=G/G_i In arbitrary product G_j, for G_j∈G_othersIf, HMSim_i,j>=ε, then by product G_jIt is incorporated in cluster, C₁=C₁∪{G_j}；

Step c：Repeat step b, until no longer have new product to be merged in cluster C₁In, thus obtaining a complete cluster C₁；

Step d：By cluster C₁In included product remove, obtain the product complete or collected works G=G/C updating₁, after updating Product complete or collected works, repeat step a to step c, until producing new cluster again, thus obtain all of cluster C={ C₁,C₂,… C_W, wherein, W≤V；

Step e：Another threshold value η is set, and wherein, η is integer, and η >=2, by included product number in all clusters Being combined of cluster less than η obtains C_small={ C₁,C₂,…,C_S, then cast out C_smallCluster in set, obtains final Gathering closes C_final=C/C_small.

Alternatively, the described division methods according to default Various Phases of Their Life Cycle determine described specified class product life Life cyclic pattern set in each lifetime value the interval position of life stage step include steps A1 with Step B1：

Step A1：The life cycle of product is corresponding by the division methods according to default Various Phases of Their Life Cycle Life stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；Step A1 concrete Step includes：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By maximum selling value institute on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Corresponding time point as the end point of introduction period, simultaneously as the starting point of growth stage,

By on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to maximum selling value as the end point of growth stage, simultaneously as the starting point in maturity period,

By on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to minimum sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By maximum selling value institute on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Corresponding time point as the end point of decline phase, simultaneously as the starting point of end period；

Step B1：Each lifetime value in described specified class product life cycle pattern is given birth to according to the said goods The division methods of life cycle stages determine each stage of the life cycle of product, so that it is determined that each lifetime value The interval position of life stage.

Alternatively, described according between default sales volume curve similarity calculating method calculate described new product sales volume curve With the similarity of each lifetime value in described lifetime value set, to obtain the step bag of similarity set Include：

Assume arbitrary life cycle mould in described new product sales volume curve and the lifetime value set of described specified class product Time series corresponding to formula is respectively：With Wherein, n >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾Product when each moment The sales volume value of product；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with Time series y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time Sequences y^baseI-th point of alignment when, definition similarity now is：

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus Obtain the similarity of lifetime value described in the lifetime value set that described new product is with described specified class product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}；

For the remaining lifetime value in the lifetime value set of described specified class product, repeat above walking Suddenly to obtain each life cycle in the sales volume curve of described new product and the lifetime value set of described specified class product The similarity of pattern, thus obtain similarity set.

Alternatively, the interval position of the described life stage according to described new product lifetime value and described relative shifting Dynamic position determines that the step of the life cycle residing for this new product includes：

For the moment t=n of new product, the described new product sales volume curve according to record and the Life Cycle of described new product The position of the relative movement of phase pattern determines that the position of two recorded curve relative movements is newly produced to obtain this moment product Product sales curve corresponds to the position t of the bass diffusion models curve of the lifetime value of described new product_bass=m, this position Put the life-cycle stages that corresponding life cycle phase is this new product；

For this new product life cycle future time instance t=n+i, i=1,2 ..., by this new product lifetime value The position t of bass diffusion models curve_bass=m+i, i=1,2 ... the life cycle phase residing for the moment is given birth to as this new product The predicted value of life cycle future time instance.

According to another aspect of the present invention, there is provided a kind of identifying device of product life cycle.

The identifying device of the product life cycle of the present invention includes：Extraction module, specifies class for obtaining in marketing system The sales data of product, then by this sales data according to the extracting method of default product life cycle pattern, default product Between the division methods in product life cycle each stage and default sales volume curve, the computational methods of similarity extract described specifying The lifetime value set of class product；Determining module, true for the division methods according to described Various Phases of Their Life Cycle The interval position of the life stage of each lifetime value in fixed described specified class product life cycle set of modes；Calculate mould Block, for for a new product under described specified class product, according to similarity calculating method between default sales volume curve Calculate the similarity of each lifetime value in described new product sales volume curve and described lifetime value set, with To similarity set；Logging modle, for the lifetime value that similarity in described similarity set is corresponding when maximum As the lifetime value of described new product, and described new product sales volume curve and described new product when recording similarity maximum The displaced posi of product lifetime value；Identification module, for the life rank according to described new product lifetime value The interval position of section and described displaced posi determine the life cycle residing for this new product.

Alternatively, described extraction module is additionally operable to：Assume to have U product in certain colony, using bass diffusion models The sales volume curve of U product is fitted obtaining collection of curves B={ B₁,B₂,...,B_U}；

The song of B in the collection of curves that division methods according to default Various Phases of Their Life Cycle obtain to described matching Line carries out the division in life cycle each stage, so that it is determined that having the product set G of complete lifecycle in U product_full= {G₁,G₂,...,G_V, wherein V≤U；

Computational methods according to similarity between default sales volume curve calculate in V product product sales volume curve two-by-two Similarity, obtains the similarity matrix between a V product, and records corresponding two of each similarity in this similarity matrix Product sales volume curve displaced posi：

One threshold epsilon of setting, is clustered to V product using this threshold value and is closed C to obtain final gathering_final；

For any one cluster C_i∈C_final, the product collection included in cluster is combined into G_i={ G₁,G₂,…,G_I, according to this In cluster, any two product is calculating the two product sales volume curve displaced posi retaining during final similarity to I product Primary products sales volume curve be overlapped merging, the product sales volume curve after being merged is：

Wherein, O_tIt is to be t in the time, the product of product sales volume curves overlapped combines, | | O_t| | represent this set unit The number of element；

For cluster C_i, bass diffusion models alignment is reused to the product sales volume curve after merging and is fitted, obtain Cluster C_iFinal product life cycle pattern P_i=(m_i,p_i,q_i)；

Belong to C to all_finalCluster all carry out as cluster C_iOperation, thus obtain specify class product product life cycle Set of modes P={ P₁,P₂,...,P_W}.

Alternatively, described extraction module is additionally operable to：Determine to have in U product according to following steps A and step B and completely give birth to The product set in life cycle：

Step A：The life cycle of product is corresponding by the division methods according to default Various Phases of Their Life Cycle Life stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period, concrete steps bag Include：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By maximum selling value institute on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Corresponding time point as the end point of introduction period, simultaneously as the starting point of growth stage,

By on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to maximum selling value as the end point of growth stage, simultaneously as the starting point in maturity period,

By on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to minimum sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By maximum selling value institute on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Corresponding time point as the end point of decline phase, simultaneously as the starting point of end period；

Step B：For the arbitrary product in U product it is assumed that its product original sales volume curve y₁={ x₁,x₂,…,x_n, Using Bass model matching curve B out₁={ b₁,b₂,…,b_n, by the curve after this matching according to the said goods Life Cycle The division methods in phase in each stage determine each stage of the life cycle of product, if matching curve out contains described product All stages of product life cycle are it is determined that this product has complete life cycle.

Alternatively, described extraction module is additionally operable to：Assume the time corresponding to any two product sales volume curves in V product Sequence is respectively：WithWherein, N >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾Product when each moment Sales volume value；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value, and in the sequence of structure, front-end and back-end all include m-1 NA value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with Time series y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time Sequences y^baseI-th point of alignment when, definition similarity now is：

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus Obtain the similarity between two sales volume curves corresponding to above-mentioned two product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}

For remaining product in V product, repeat the above steps, thus it is bent to obtain in V product product sales volume two-by-two The similarity of line.

Alternatively, described extraction module is additionally operable to：Obtain final gathering according to following steps and close C_final：For having had The product complete or collected works G={ G of whole life cycle₁,G₂,…G_v}：

Step a：A product G is chosen from described product complete or collected works G_i, as first cluster C₁={ G_iIn product；

Step b：By G_iRemove from product complete or collected works, take out and in product complete or collected works G, remove G_iSurplus products G_others=G/G_i In arbitrary product G_j, for G_j∈G_othersIf, HMSim_i,j>=ε, then by product G_jIt is incorporated in cluster, C₁=C₁∪{G_j}；

Step c：Repeat step b, until no longer have new product to be merged in cluster C₁In, thus obtaining a complete cluster C₁；

Step d：By cluster C₁In included product remove, obtain the product complete or collected works G=G/C updating₁, after updating Product complete or collected works, repeat step a to step c, until producing new cluster again, thus obtain all of cluster C={ C₁,C₂,… C_W, wherein, W≤V；

Step e：Another threshold value η is set, and wherein, η is integer, and η >=2, by included product number in all clusters Being combined of cluster less than η obtains C_small={ C₁,C₂,…,C_S, then cast out C_smallCluster in set, obtains final Gathering closes C_final=C/C_small.

Alternatively, described determining module is additionally operable to：Determine described specified class product life according to following steps A1 and step B1 The interval position of the life stage of each lifetime value in life cyclic pattern set：

Step A1：The life cycle of product is corresponding by the division methods according to default Various Phases of Their Life Cycle Life stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；Step A1 concrete Step includes：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By maximum selling value institute on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Corresponding time point as the end point of introduction period, simultaneously as the starting point of growth stage,

By on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to maximum selling value as the end point of growth stage, simultaneously as the starting point in maturity period,

By on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to minimum sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By maximum selling value institute on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Corresponding time point as the end point of decline phase, simultaneously as the starting point of end period；

Step B1：Each lifetime value in described specified class product life cycle pattern is given birth to according to the said goods The division methods of life cycle stages determine each stage of the life cycle of product, so that it is determined that each lifetime value The interval position of life stage.

Alternatively, described computing module is additionally operable to：

Assume arbitrary life cycle mould in described new product sales volume curve and the lifetime value set of described specified class product Time series corresponding to formula is respectively：With Wherein, n >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾Product when each moment The sales volume value of product；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with Time series y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time Sequences y^baseI-th point of alignment when, definition similarity now is：

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus Obtain the similarity of lifetime value described in the lifetime value set that described new product is with described specified class product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}；

For the remaining lifetime value in the lifetime value set of described specified class product, repeat above walking Suddenly to obtain each life cycle in the sales volume curve of described new product and the lifetime value set of described specified class product The similarity of pattern, thus obtain similarity set.

Alternatively, described identification module is additionally operable to：For the moment t=n of new product, according to the described new product pin of record Amount curve determines two recorded curve relative movements with the position of the relative movement of lifetime value of described new product Position with obtain this moment product new product sales curve correspond to described new product lifetime value Bath diffusion The position t of model curve_bass=m, the life cycle phase corresponding to this position is the product life cycle rank of this new product Section；

For this new product life cycle future time instance t=n+i, i=1,2 ..., by this new product lifetime value The position t of bass diffusion models curve_bass=m+i, i=1,2 ... the life cycle phase residing for the moment is given birth to as this new product The predicted value of life cycle future time instance.

Technology according to the present invention scheme, due to can be according to the mathematics division side of given life-cycle stages Method divides to life-cycle stages, and the life cycle phase simultaneously a product being presently in is identified and not Carry out moment life cycle to be predicted, not only avoid the impact of subjective factor, and for new product without utilizing model Carry out reevaluating matching to it, amount of calculation is relatively small, finally improve accuracy rate and the efficiency of product life cycle identification.

Brief description

Accompanying drawing is used for more fully understanding the present invention, does not constitute inappropriate limitation of the present invention.Wherein：

Fig. 1 is a kind of schematic diagram of the recognition methods of product life cycle according to embodiments of the present invention；

Fig. 2A, Fig. 2 B and Fig. 2 C is the schematic diagram of Similarity Measure process according to embodiments of the present invention；

Fig. 3 is the schematic diagram of the process that commodity curve merges in cluster according to embodiments of the present invention；

Fig. 4 is a kind of schematic diagram of the identifying device of product life cycle according to embodiments of the present invention.

Specific embodiment

Below in conjunction with accompanying drawing, the one exemplary embodiment of the present invention is explained, various including the embodiment of the present invention Details is to help understanding it should they are thought only exemplary.Therefore, those of ordinary skill in the art it should be noted that Arrive, various changes and modifications can be made to the embodiments described herein, without departing from scope and spirit of the present invention.With Sample, for clarity and conciseness, eliminates the description to known function and structure in description below.

Fig. 1 is a kind of schematic diagram of the recognition methods of product life cycle according to embodiments of the present invention.As shown in figure 1, The recognition methods of the product life cycle of the present invention mainly includes steps S10 to S14.

Step S10：Obtain the sales data specifying class product in marketing system, then by this sales data according to default The extracting method of product life cycle pattern, the division methods of default Various Phases of Their Life Cycle and default sales volume Between curve, the computational methods of similarity extract the lifetime value set specifying class product.In step slo, sales volume curve Refer to record product selling time and the curve of the sales volume corresponding to each time, the product being previously mentioned possesses following spy Levy：

(1) life cycle characteristic of product is represented using the sales volume of product, is designated as y=f (t), and wherein t is the record time, And time interval is for fixed value (for example：Week or the moon).

(2) product itself has and can be used for the description information of class discrimination (for example：Brand, category etc.) it is assumed that a product Product Good_i, it represents the 4G cell phone of XXX brand, then its brand (XXX) and category (4G cell phone) are and are used for retouching of class discrimination State information.

(3) for record time t=1, it is not offered as the starting point of product whole life cycle, only represent and can get this product First time point of product sales volume value, that is, as t=1, a product is likely to be at any life cycle phase.

For step S10, after getting the sales data of specified class product from marketing system, by following concrete step The rapid lifetime value extracting specified class product：Assume to have U product in certain colony, using bass diffusion models to U The sales volume curve of individual product is fitted obtaining collection of curves B={ B₁,B₂,...,B_U}.For the parameter in bass diffusion models M, p, q, embodiment of the present invention technical scheme carries out parameter Estimation using Levenberg-Marquardt algorithm.Levenberg- Marquardt algorithm (also referred to as damped least-squares, DLS), is that one kind is used for solving non-linear least square asking The algorithm of topic, is usually used in the matching of nonlinear curve.

Step S101：The curve set that division methods according to default Various Phases of Their Life Cycle obtain to described matching In conjunction, the curve of B carries out the division in life cycle each stage, so that it is determined that having the product collection of complete lifecycle in U product Close G_full={ G₁,G₂,...,G_V, wherein V≤U.In step S101, mainly include steps A and step B：

Step A：The life cycle of product is corresponding by the division methods according to default Various Phases of Their Life Cycle Life stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；Step A concrete Step includes：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By maximum selling value institute on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Corresponding time point as the end point of introduction period, simultaneously as the starting point of growth stage,

By on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to maximum selling value as the end point of growth stage, simultaneously as the starting point in maturity period,

By on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to minimum sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By maximum selling value institute on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Corresponding time point as the end point of decline phase, simultaneously as the starting point of end period；

Step B：For the arbitrary product in U product it is assumed that its product original sales volume curve y₁={ x₁,x₂,…,x_n, Using Bass model matching curve B out₁={ b₁,b₂,…,b_n, by the curve after this matching according to the said goods Life Cycle The division methods in phase in each stage determine each stage of the life cycle of product, if matching curve out contains described product All stages of product life cycle are it is determined that this product has complete life cycle.

Step S102：Computational methods according to similarity between default sales volume curve calculate product two-by-two in V product The similarity of sales volume curve, obtains the similarity matrix between a V product, and records each similarity in this similarity matrix Corresponding two product sales volume curve displaced posi.In step s 102, calculate product sales volume curve two-by-two in V product The process of similarity be：

Assume that in V product, the time series corresponding to any two product sales volume curves is respectively：WithWherein, n >=m, table respectively Show sequences y⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾The sales volume value of the product when each moment；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value, and in the sequence of structure, front-end and back-end all include m-1 NA value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with Time series y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time Sequences y^baseI-th point of alignment when, definition similarity now is：

Wherein, for any one situation of i, the Part I of similarity calculates time series y⁽²⁾With time series y^baseCosine similarity between the set of point of the intersection of non-null value.In order to avoid certain seasonal effect in time series of intersection Value is zero situation, and technical solution of the present invention employs the thought that Laplce smooths, in the cosine phase calculating intersection When seemingly spending, the value for each data point is carried out plus a process；When Part II is to calculate similarity, due to not using all counting Introduce according to (in sliding process merely with intersection calculate cosine similarity) and two length of time series different (n ≠ m) Penalty term, wherein λ be penalty coefficient.

Different situations in 3 of i in defining for similarity, correspond respectively to Fig. 2A, Fig. 2 B and Fig. 2 C；Similarity meter As shown in Fig. 2 wherein, imaginary curve is time series y to calculation process^base(NA value is not drawn in fig. 2), solid-line curve is time series y⁽²⁾, two red vertical dotted line are the data (i.e. the intersection data of imaginary curve and solid-line curve) participating in Similarity Measure.

By calculating time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, Thus obtaining the similarity between two sales volume curves corresponding to above-mentioned two product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}

For remaining product in V product, repeat the above steps, thus it is bent to obtain in V product product sales volume two-by-two The similarity of line, thus obtain similarity matrix

Due to HMSim_u,vAnd HMSim_v,uValue is identical, and therefore similarity matrix is symmetrical matrix, repeats numerical value in above-mentioned square Do not show in battle array.

Step S103：One threshold epsilon of setting, is clustered to V product using this threshold value and is closed with obtaining final gathering C_final.In step s 103, using this threshold value, V product is clustered and closed C to obtain final gathering_finalConcrete bag Include step a to step e：

For the product complete or collected works G={ G with complete lifecycle₁,G₂,…G_v}：

Step a：A product G is chosen from described product complete or collected works G_i, as first cluster C₁={ G_iIn product；

Step b：By G_iRemove from product complete or collected works, take out and in product complete or collected works G, remove G_iSurplus products G_others=G/G_i In arbitrary product G_j, for G_j∈G_othersIf, HMSim_i,j>=ε, then by product G_jIt is incorporated in cluster, C₁=C₁∪{G_j}；

Step c：Repeat step b, until no longer have new product to be merged in cluster C₁In, thus obtaining a complete cluster C₁；

Step d：By cluster C₁In included product remove, obtain the product complete or collected works G=G/C updating₁, after updating Product complete or collected works, repeat step a to step c, until producing new cluster again, thus obtain all of cluster C={ C₁,C₂,… C_W, wherein, W≤V；

Step e：Another threshold value η is set, and wherein, η is integer, and η >=2, by included product number in all clusters Being combined of cluster less than η obtains C_small={ C₁,C₂,…,C_S, then cast out C_smallCluster in set, obtains final Gathering closes C_final=C/C_small.

Step S104：For any one cluster C_i∈C_final, the product collection included in cluster is combined into G_i={ G₁,G₂,…, G_I, the two product sales volume curve relative movement positions retaining during final similarity are being calculated according to any two product in this cluster Put the primary products sales volume curve to I product to be overlapped merging, the product sales volume curve after being merged is：

Wherein, O_tIt is to be t in the time, the product of product sales volume curves overlapped combines, | | O_t| | represent this set unit The number of element；The process example merging is as shown in figure 3, in the curve of 4 shown in Fig. 3,3 solid lines are 3 products in a cluster The sales volume curve (being alignd according to displaced posi when calculating similarity before) of product, dotted line is the product pin after merging Amount curve.

Step S105：For cluster C_i, bass diffusion models alignment is reused to the product sales volume curve after merging and carries out Matching, obtains cluster C_iFinal product life cycle pattern P_i=(m_i,p_i,q_i).

Step S106：Belong to C to all_finalCluster all carry out as cluster C_iOperation, thus obtain specify class product product Product lifetime value set P={ P₁,P₂,...,P_W}.

Step S11：Determine the described specified week class life of product according to the division methods of described Various Phases of Their Life Cycle The interval position of the life stage of each lifetime value in phase set of modes.In this step, the following step of main inclusion Rapid A1 and step B1：

Step A1：The life cycle of product is corresponding by the division methods according to default Various Phases of Their Life Cycle Life stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；Step A1 concrete Step includes：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By maximum selling value institute on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Corresponding time point as the end point of introduction period, simultaneously as the starting point of growth stage,

By on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to maximum selling value as the end point of growth stage, simultaneously as the starting point in maturity period,

By on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to minimum sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By maximum selling value institute on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Corresponding time point as the end point of decline phase, simultaneously as the starting point of end period；

Step B1：Each lifetime value in described specified class product life cycle pattern is given birth to according to the said goods The division methods of life cycle stages determine each stage of the life cycle of product, so that it is determined that each lifetime value The interval position of life stage.

Step S12：For a new product under described specified class product, according to similarity between default sales volume curve Computational methods calculate the phase of each lifetime value in described new product sales volume curve and described lifetime value set Like spending, to obtain similarity set.In this step, specifically include the steps：

Assume arbitrary life cycle mould in described new product sales volume curve and the lifetime value set of described specified class product Time series corresponding to formula is respectively：With Wherein, n >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾Product when each moment The sales volume value of product；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with Time series y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time Sequences y^baseI-th point of alignment when, definition similarity now is：

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus Obtain the similarity of lifetime value described in the lifetime value set that described new product is with described specified class product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}；

For the remaining lifetime value in the lifetime value set of described specified class product, repeat above walking Suddenly to obtain each life cycle in the sales volume curve of described new product and the lifetime value set of described specified class product The similarity of pattern, thus obtain similarity set.

Step S13：When will be maximum for similarity in described similarity set, corresponding lifetime value be as described new The lifetime value of product, and described new product sales volume curve and described new product life cycle when recording similarity maximum The displaced posi of pattern.

Step S14：The interval position of the life stage according to described new product lifetime value and described relative movement Position determines the life cycle residing for this new product.In this step, for the moment t=n of new product, according to record New product sales volume curve determines two recorded curves with the position of the relative movement of lifetime value of described new product The position of relative movement is to obtain the lifetime value that this moment product new product sales curve corresponds to described new product The position t of bass diffusion models curve_bass=m, the life cycle phase corresponding to this position is the product life of this new product The life phase of the cycles；

For this new product life cycle future time instance t=n+i, i=1,2 ..., by this new product lifetime value The position t of bass diffusion models curve_bass=m+i, i=1,2 ... the life cycle phase residing for the moment is given birth to as this new product The predicted value of life cycle future time instance.

Fig. 4 is a kind of schematic diagram of the identifying device of product life cycle according to embodiments of the present invention.The product of the present invention The identifying device 40 of product life cycle mainly include extraction module 41, determining module 42, computing module 43, logging modle 44, with And identification module 45；Extraction module 41 is used for obtaining the sales data specifying class product in marketing system, then by this sale number According to according to the extracting method of default product life cycle pattern, the division methods of default Various Phases of Their Life Cycle, with And the computational methods of similarity extract the lifetime value set specifying class product between default sales volume curve；Determining module 42 are used for determining described specified class product life cycle set of modes according to the division methods of described Various Phases of Their Life Cycle In each lifetime value life stage interval position；Computing module 43 is used for under described specified class product Individual new product, calculates described new product sales volume curve and described life according to similarity calculating method between default sales volume curve The similarity of each lifetime value in cyclic pattern set, to obtain similarity set；Logging modle 44 is used for institute State corresponding lifetime value during similarity maximum in similarity set as the lifetime value of described new product, and And record similarity maximum when described new product sales volume curve and described new product lifetime value displaced posi；Know Other module 45 is used for the interval position of the life stage according to described new product lifetime value and described displaced posi Determine the life cycle residing for this new product.

Extraction module 41 can be additionally used in：Assume to have U product in certain colony, using bass diffusion models, U is produced The sales volume curve of product is fitted obtaining collection of curves B={ B₁,B₂,...,B_U}；According to default Various Phases of Their Life Cycle Division methods collection of curves that described matching is obtained in the curve of B carry out the division in life cycle each stage, so that it is determined that U There is in individual product the product set G of complete lifecycle_full={ G₁,G₂,...,G_V, wherein V≤U；According to default sales volume Between curve, the computational methods of similarity calculate the similarity of product sales volume curve two-by-two in V product, obtain between a V product Similarity matrix, and record the corresponding two product sales volume curve relative movement positions of each similarity in this similarity matrix Put：

One threshold epsilon of setting, is clustered to V product using this threshold value and is closed C to obtain final gathering_final；

For any one cluster C_i∈C_final, the product collection included in cluster is combined into G_i={ G₁,G₂,…,G_I, according to this In cluster, any two product is calculating the two product sales volume curve displaced posi retaining during final similarity to I product Primary products sales volume curve be overlapped merging, the product sales volume curve after being merged is：

Wherein, O_tIt is to be t in the time, the product of product sales volume curves overlapped combines, | | O_t| | represent this set unit The number of element；

For cluster C_i, bass diffusion models alignment is reused to the product sales volume curve after merging and is fitted, obtain Cluster C_iFinal product life cycle pattern P_i=(m_i,p_i,q_i)；

Belong to C to all_finalCluster all carry out as cluster C_iOperation, thus obtain specify class product product life cycle Set of modes P={ P₁,P₂,...,P_W}.

Extraction module 41 can be additionally used in：Determine in U product that there is complete lifecycle according to following steps A and step B Product set：

Step A：The life cycle of product is corresponding by the division methods according to default Various Phases of Their Life Cycle Life stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period, concrete steps bag Include：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By maximum selling value institute on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Corresponding time point as the end point of introduction period, simultaneously as the starting point of growth stage,

By on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to maximum selling value as the end point of growth stage, simultaneously as the starting point in maturity period,

By on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to minimum sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By maximum selling value institute on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Corresponding time point as the end point of decline phase, simultaneously as the starting point of end period；

Step B：For the arbitrary product in U product it is assumed that its product original sales volume curve y₁={ x₁,x₂,…,x_n, Using Bass model matching curve B out₁={ b₁,b₂,…,b_n, by the curve after this matching according to the said goods Life Cycle The division methods in phase in each stage determine each stage of the life cycle of product, if matching curve out contains described product All stages of product life cycle are it is determined that this product has complete life cycle.

Extraction module 41 can be additionally used in：Assume that in V product, the time series corresponding to any two product sales volume curves is divided It is not：WithWherein, n >=m, Represent sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾The sales volume of the product when each moment Value；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value, and in the sequence of structure, front-end and back-end all include m-1 NA value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with Time series y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time Sequences y^baseI-th point of alignment when, definition similarity now is：

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus Obtain the similarity between two sales volume curves corresponding to above-mentioned two product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}

For remaining product in V product, repeat the above steps, thus it is bent to obtain in V product product sales volume two-by-two The similarity of line.

Extraction module 41 can be additionally used in：Obtain final gathering according to following steps and close C_final：For having complete life The product complete or collected works G={ G in cycle₁,G₂,…G_v}：

Step a：A product G is chosen from described product complete or collected works G_i, as first cluster C₁={ G_iIn product；

Step b：By G_iRemove from product complete or collected works, take out and in product complete or collected works G, remove G_iSurplus products G_others=G/G_i In arbitrary product G_j, for G_j∈G_othersIf, HMSim_i,j>=ε, then by product G_jIt is incorporated in cluster, C₁=C₁∪{G_j}；

Step c：Repeat step b, until no longer have new product to be merged in cluster C₁In, thus obtaining a complete cluster C₁；

Step d：By cluster C₁In included product remove, obtain the product complete or collected works G=G/C updating₁, after updating Product complete or collected works, repeat step a to step c, until producing new cluster again, thus obtain all of cluster C={ C₁,C₂,… C_W, wherein, W≤V；

Step e：Another threshold value η is set, and wherein, η is integer, and η >=2, by included product number in all clusters Being combined of cluster less than η obtains C_small={ C₁,C₂,…,C_S, then cast out C_smallCluster in set, obtains final Gathering closes C_final=C/C_small.

Determining module 42 can be additionally used in：Determine described specified class product life cycle mould according to following steps A1 and step B1 The interval position of the life stage of each lifetime value in formula set：

Step A1：The life cycle of product is corresponding by the division methods according to default Various Phases of Their Life Cycle Life stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；Step A1 concrete Step includes：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By maximum selling value institute on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Corresponding time point as the end point of introduction period, simultaneously as the starting point of growth stage,

By on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to maximum selling value as the end point of growth stage, simultaneously as the starting point in maturity period,

By on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to minimum sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By maximum selling value institute on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Corresponding time point as the end point of decline phase, simultaneously as the starting point of end period；

Step B1：Each lifetime value in described specified class product life cycle pattern is given birth to according to the said goods The division methods of life cycle stages determine each stage of the life cycle of product, so that it is determined that each lifetime value The interval position of life stage.

Computing module 43 can be additionally used in：Assume described new product sales volume curve and the life cycle mould of described specified class product In formula set, the time series corresponding to arbitrary lifetime value is respectively： WithWherein, n >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mPoint Wei not sequences y⁽¹⁾And y⁽²⁾The sales volume value of the product when each moment；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with Time series y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time Sequences y^baseI-th point of alignment when, definition similarity now is：

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus Obtain the similarity of lifetime value described in the lifetime value set that described new product is with described specified class product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}；

For the remaining lifetime value in the lifetime value set of described specified class product, repeat above walking Suddenly to obtain each life cycle in the sales volume curve of described new product and the lifetime value set of described specified class product The similarity of pattern, thus obtain similarity set.

Identification module 45 can be additionally used in：For the moment t=n of new product, according to the described new product sales volume curve of record Determine the position of two recorded curve relative movements with the position of the relative movement of the lifetime value of described new product To obtain the bass diffusion models song that this moment product new product sales curve corresponds to the lifetime value of described new product The position t of line_bass=m, the life cycle phase corresponding to this position is the life-cycle stages of this new product；For This new product life cycle future time instance t=n+i, i=1,2 ..., by the bass diffusion models of this new product lifetime value The position t of curve_bass=m+i, i=1,2 ... the life cycle phase residing for the moment as this new product life cycle following when The predicted value carved.

Technical scheme according to embodiments of the present invention, due to can be according to the mathematics of given life-cycle stages Division methods divide to life-cycle stages, and the life cycle phase simultaneously a product being presently in is known It is not predicted with future time instance life cycle, not only avoid the impact of subjective factor, and for new product without profit Carry out reevaluating matching with model to it, amount of calculation is relatively small, finally improve the accuracy rate of product life cycle identification And efficiency.

Above-mentioned specific embodiment, does not constitute limiting the scope of the invention.Those skilled in the art should be bright White, depending on design requirement and other factors, various modifications, combination, sub-portfolio and replacement can occur.Any Modification, equivalent and improvement of being made within the spirit and principles in the present invention etc., should be included in the scope of the present invention Within.

Claims (16)

1. a kind of recognition methods of product life cycle is it is characterised in that include：

Obtain the sales data specifying class product in marketing system, then by this sales data according to the default product life cycle Similar between the extracting method of pattern, the division methods of default Various Phases of Their Life Cycle and default sales volume curve The computational methods of degree extract the lifetime value set of described specified class product；

Determined in described specified class product life cycle set of modes according to the division methods of described Various Phases of Their Life Cycle The interval position of the life stage of each lifetime value；

For a new product under described specified class product, calculated according to similarity calculating method between default sales volume curve The similarity of each lifetime value in described new product sales volume curve and described lifetime value set, to obtain phase Like degree set；

When will be maximum for similarity in described similarity set, corresponding lifetime value be as the Life Cycle of described new product Phase pattern, and record similarity maximum when described new product sales volume curve and described new product lifetime value relative shifting Dynamic position；

The interval position of the life stage according to described new product lifetime value and described displaced posi determine that this is new Life cycle residing for product.

2. method according to claim 1 is it is characterised in that the lifetime value set of class product is specified in described extraction Step include：

Assume to have U product in certain colony, using bass diffusion models, the sales volume curve of U product is fitted obtaining Collection of curves B={ B₁,B₂,...,B_U}；

In the collection of curves that division methods according to default Various Phases of Their Life Cycle obtain to described matching, the curve of B enters The division in row life cycle each stage, so that it is determined that have the product set G of complete lifecycle in U product_full={ G₁, G₂,...,G_V, wherein V≤U；

Computational methods according to similarity between default sales volume curve calculate the similar of product sales volume curve two-by-two in V product Degree, obtains the similarity matrix between a V product, and records corresponding two products of each similarity in this similarity matrix Sales volume curve displaced posi：

$M_{HMSim}=\left[\begin{array}{ccccc}-& {\mathrm{HMSim}}_{1,2}& {\mathrm{HMSim}}_{1,3}& ...& {\mathrm{HMSim}}_{1,V}\\ -& -& {\mathrm{HMSim}}_{2,3}& ...& {\mathrm{HMSim}}_{2,V}\\ -& -& -& ...& ...\\ -& -& -& ...& {\mathrm{HMSim}}_{V-1,V}\\ -& -& -& -& -\end{array}\right]$

One threshold epsilon of setting, is clustered to V product using this threshold value and is closed C to obtain final gathering_final；

For any one cluster C_i∈C_final, the product collection included in cluster is combined into G_i={ G₁,G₂,…,G_I, appoint according in this cluster Two product sales volume curve displaced posi that two products of meaning retain in the final similarity of calculating are original to I product Product sales volume curve is overlapped merging, and the product sales volume curve after being merged is：

$y_i^c=f\left(t\right)=\frac{\underset{j\∈O_t}{\Σ}{G}_j}{\left|\right|O_t\left|\right|}$

Wherein, O_tIt is to be t in the time, the product of product sales volume curves overlapped combines, | | O_t| | represent this set element Number；

For cluster C_i, bass diffusion models alignment is reused to the product sales volume curve after merging and is fitted, obtain cluster C_i? Whole product life cycle pattern P_i=(m_i, p_i, q_i)；

Belong to C to all_finalCluster all carry out as cluster C_iOperation, thus obtain specify class product product life cycle pattern Set P={ P₁,P₂,...,P_W}.

3. method according to claim 2 it is characterised in that described according to default Various Phases of Their Life Cycle draw In point collection of curves that matching described in method obtains, the curve of B carries out the division in life cycle each stage, so that it is determined that U product In there is complete lifecycle the step of product set include steps A and step B：

Step A：The life cycle of product is corresponding life by the division methods according to default Various Phases of Their Life Cycle In the stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；The concrete steps of step A Including：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By corresponding to maximum selling value on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Time point as the introduction period end point, simultaneously as the starting point of growth stage,

By the maximum on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to sales volume value as the end point of growth stage, simultaneously as the starting point in maturity period,

By the minimum on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By corresponding to maximum selling value on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Time point as the decline phase end point, simultaneously as the starting point of end period；

Step B：For the arbitrary product in U product it is assumed that its product original sales volume curve y₁={ x₁,x₂,…,x_n, utilize Bass model matching curve B out₁={ b₁,b₂,…,b_n, will be each according to the said goods life cycle for the curve after this matching The division methods in stage determine each stage of the life cycle of product, if matching curve out contains described product life All stages in life cycle are it is determined that this product has complete life cycle.

4. method according to claim 2 is it is characterised in that the described meter according to similarity between default sales volume curve The step that calculation method calculates the similarity of product sales volume curve two-by-two in V product includes：

Assume that in V product, the time series corresponding to any two product sales volume curves is respectively： WithWherein, n >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mRespectively For sequences y⁽¹⁾And y⁽²⁾The sales volume value of the product when each moment；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value, and in the sequence of structure, front-end and back-end all include m-1 NA value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with time sequence Row y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time series y^base I-th point of alignment when, definition similarity now is：

${\mathrm{Sim}}_i=\left\{\begin{array}{cc}\frac{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}(y_{m+j-1}^{base}+1)\&CenterDot;(y_{m-j+1}^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}{(y_{m+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}{(y_{m-j+1}^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2i}{m+n},& 0<i\&le;m-1\\ \frac{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}(y_{i+j-1}^{base}+1)\&CenterDot;(y_j^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{(y_{i+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{(y_j^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2m}{m+n},& m-1<i\&le;n-1\\ \frac{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}(y_{i+j-1}^{base}+1)\&CenterDot;(y_j^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}{(y_{i+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}{(y_j^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2(n+m-1-i)}{m+n},& n-1<i\&le;n+m-1\end{array}\right.$

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus obtaining The similarity between two sales volume curves corresponding to above-mentioned two product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}

For remaining product in V product, repeat the above steps, thus obtaining in V product product sales volume curve two-by-two Similarity.

5. method according to claim 2 is it is characterised in that described clustered to V product using this threshold value to obtain Close C to final gathering_finalStep include：

For the product complete or collected works G={ G with complete lifecycle₁,G₂,…G_v}：

Step a：A product G is chosen from described product complete or collected works G_i, as first cluster C₁={ G_iIn product；

Step b：By G_iRemove from product complete or collected works, take out and in product complete or collected works G, remove G_iSurplus products G_others=G/G_iIn Arbitrary product G_j, for G_j∈G_othersIf, HMSim_i,j>=ε, then by product G_jIt is incorporated in cluster, C₁=C₁∪{G_j}；

Step c：Repeat step b, until no longer have new product to be merged in cluster C₁In, thus obtaining a complete cluster C₁；

Step d：By cluster C₁In included product remove, obtain the product complete or collected works G=G/C updating₁, for the product after updating Product complete or collected works, repeat step a to step c, until producing new cluster again, thus obtain all of cluster C={ C₁,C₂,…C_W, Wherein, W≤V；

Step e：Another threshold value η is set, and wherein, η is integer, and η >=2, and included product number in all clusters is less than η Being combined of cluster obtain C_small={ C₁,C₂,…,C_S, then cast out C_smallCluster in set, obtains final gathering and closes C_final=C/C_small.

6. method according to claim 1 it is characterised in that described according to default Various Phases of Their Life Cycle draw Point method determines the interval position of the life stage of each lifetime value in described specified class product life cycle set of modes The step put includes steps A1 and step B1：

Step A1：The life cycle of product is corresponding life by the division methods according to default Various Phases of Their Life Cycle In the stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；The concrete steps of step A1 Including：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By corresponding to maximum selling value on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Time point as the introduction period end point, simultaneously as the starting point of growth stage,

By the maximum on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to sales volume value as the end point of growth stage, simultaneously as the starting point in maturity period,

By the minimum on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By corresponding to maximum selling value on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Time point as the decline phase end point, simultaneously as the starting point of end period；

Step B1：By each lifetime value in described specified class product life cycle pattern according to the said goods Life Cycle The division methods in phase in each stage determine each stage of the life cycle of product, so that it is determined that the life of each lifetime value The interval position in stage.

7. method according to claim 1 it is characterised in that described according to Similarity Measure between default sales volume curve Method calculates the similarity of each lifetime value in described new product sales volume curve and described lifetime value set, Included with the step obtaining similarity set：

Assume that in described new product sales volume curve and the lifetime value set of described specified class product, arbitrary lifetime value institute is right The time series answered is respectively：With Wherein, n >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾Product when each moment The sales volume value of product；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with time sequence Row y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time series y^base I-th point of alignment when, definition similarity now is：

${\mathrm{Sim}}_i=\left\{\begin{array}{cc}\frac{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}(y_{m+j-1}^{base}+1)\&CenterDot;(y_{m-j+1}^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}{(y_{m+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}{(y_{m-j+1}^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2i}{m+n},& 0<i\&le;m-1\\ \frac{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}(y_{i+j-1}^{base}+1)\&CenterDot;(y_j^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{(y_{i+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{(y_j^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2m}{m+n},& m-1<i\&le;n-1\\ \frac{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}(y_{i+j-1}^{base}+1)\&CenterDot;(y_j^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}{(y_{i+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}{(y_j^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2(n+m-1-i)}{m+n},& n-1<i\&le;n+m-1\end{array}\right.$

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus obtaining The similarity of lifetime value described in the lifetime value set with described specified class product for the described new product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}；

For the remaining lifetime value in the lifetime value set of described specified class product, repeat above step with Obtain each lifetime value in the sales volume curve of described new product and the lifetime value set of described specified class product Similarity, thus obtaining similarity set.

8. method according to claim 1 is it is characterised in that the described life according to described new product lifetime value The interval position in stage and described displaced posi determine that the step of the life cycle residing for this new product includes：

For the moment t=n of new product, the described new product sales volume curve according to record and the life cycle mould of described new product The position of the relative movement of formula determines the position of two recorded curve relative movements to obtain this moment product new product pin Sell the position t of the bass diffusion models curve of the lifetime value that curve corresponds to described new product_bass=m, this position institute Corresponding life cycle phase is the life-cycle stages of this new product；

For this new product life cycle future time instance t=n+i, i=1,2 ..., by the Bath of this new product lifetime value The position t of diffusion model curve_bass=m+i, i=1,2 ... the life cycle phase residing for the moment is as this new product Life Cycle The predicted value of phase future time instance.

9. a kind of identifying device of product life cycle is it is characterised in that include：

Extraction module, for obtaining the sales data specifying class product in marketing system, then by this sales data according to default The extracting method of product life cycle pattern, the division methods of default Various Phases of Their Life Cycle and default pin Between amount curve, the computational methods of similarity extract the lifetime value set of described specified class product；

Determining module, for determining the described specified week class life of product according to the division methods of described Various Phases of Their Life Cycle The interval position of the life stage of each lifetime value in phase set of modes；

Computing module, for for a new product under described specified class product, according to similar between default sales volume curve Degree computational methods calculate each lifetime value in described new product sales volume curve and described lifetime value set Similarity, to obtain similarity set；

Logging modle, for will be maximum for similarity in described similarity set when corresponding lifetime value as described new The lifetime value of product, and described new product sales volume curve and described new product life cycle when recording similarity maximum The displaced posi of pattern；

Identification module, the interval position for the life stage according to described new product lifetime value and described relative movement Position determines the life cycle residing for this new product.

10. device according to claim 9 is it is characterised in that described extraction module is additionally operable to：

Assume to have U product in certain colony, using bass diffusion models, the sales volume curve of U product is fitted obtaining Collection of curves B={ B₁,B₂,...,B_U}；

In the collection of curves that division methods according to default Various Phases of Their Life Cycle obtain to described matching, the curve of B enters The division in row life cycle each stage, so that it is determined that have the product set G of complete lifecycle in U product_full={ G₁, G₂,...,G_V, wherein V≤U；

Computational methods according to similarity between default sales volume curve calculate the similar of product sales volume curve two-by-two in V product Degree, obtains the similarity matrix between a V product, and records corresponding two products of each similarity in this similarity matrix Sales volume curve displaced posi：

$M_{HMSim}=\left[\begin{array}{ccccc}-& {\mathrm{HMSim}}_{1,2}& {\mathrm{HMSim}}_{1,3}& ...& {\mathrm{HMSim}}_{1,V}\\ -& -& {\mathrm{HMSim}}_{2,3}& ...& {\mathrm{HMSim}}_{2,V}\\ -& -& -& ...& ...\\ -& -& -& ...& {\mathrm{HMSim}}_{V-1,V}\\ -& -& -& -& -\end{array}\right]$

One threshold epsilon of setting, is clustered to V product using this threshold value and is closed C to obtain final gathering_final；

For any one cluster C_i∈C_final, the product collection included in cluster is combined into G_i={ G₁,G₂,…,G_I, appoint according in this cluster Two product sales volume curve displaced posi that two products of meaning retain in the final similarity of calculating are original to I product Product sales volume curve is overlapped merging, and the product sales volume curve after being merged is：

$y_i^c=f\left(t\right)=\frac{\underset{j\&Element;O_t}{\&Sigma;}{G}_j}{\left|\right|O_t\left|\right|}$

Wherein, O_tIt is to be t in the time, the product of product sales volume curves overlapped combines, | | O_t| | represent this set element Number；

For cluster C_i, bass diffusion models alignment is reused to the product sales volume curve after merging and is fitted, obtain cluster C_i? Whole product life cycle pattern P_i=(m_i, p_i, q_i)；

Belong to C to all_finalCluster all carry out as cluster C_iOperation, thus obtain specify class product product life cycle pattern Set P={ P₁,P₂,...,P_W}.

11. devices according to claim 10 are it is characterised in that described extraction module is additionally operable to：

Determine the product set in U product with complete lifecycle according to following steps A and step B：

Step A：The life cycle of product is corresponding life by the division methods according to default Various Phases of Their Life Cycle In the stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period, concrete steps include：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By corresponding to maximum selling value on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Time point as the introduction period end point, simultaneously as the starting point of growth stage,

By the maximum on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to sales volume value as the end point of growth stage, simultaneously as the starting point in maturity period,

By the minimum on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By corresponding to maximum selling value on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Time point as the decline phase end point, simultaneously as the starting point of end period；

Step B：For the arbitrary product in U product it is assumed that its product original sales volume curve y₁={ x₁,x₂,…,x_n, utilize Bass model matching curve B out₁={ b₁,b₂,…,b_n, will be each according to the said goods life cycle for the curve after this matching The division methods in stage determine each stage of the life cycle of product, if matching curve out contains described product life All stages in life cycle are it is determined that this product has complete life cycle.

12. devices according to claim 10 are it is characterised in that described extraction module is additionally operable to：

Assume that in V product, the time series corresponding to any two product sales volume curves is respectively： WithWherein, n >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively Sequences y⁽¹⁾And y⁽²⁾The sales volume value of the product when each moment；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value, and in the sequence of structure, front-end and back-end all include m-1 NA value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with time sequence Row y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time series y^base I-th point of alignment when, definition similarity now is：

${\mathrm{Sim}}_i=\left\{\begin{array}{cc}\frac{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}(y_{m+j-1}^{base}+1)\&CenterDot;(y_{m-j+1}^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}{(y_{m+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}{(y_{m-j+1}^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2i}{m+n},& 0<i\&le;m-1\\ \frac{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}(y_{i+j-1}^{base}+1)\&CenterDot;(y_j^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{(y_{i+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{(y_j^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2m}{m+n},& m-1<i\&le;n-1\\ \frac{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}(y_{i+j-1}^{base}+1)\&CenterDot;(y_j^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}{(y_{i+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}{(y_j^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2(n+m-1-i)}{m+n},& n-1<i\&le;n+m-1\end{array}\right.$

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus obtaining The similarity between two sales volume curves corresponding to above-mentioned two product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}

For remaining product in V product, repeat the above steps, thus obtaining in V product product sales volume curve two-by-two Similarity.

13. devices according to claim 10 are it is characterised in that described extraction module is additionally operable to：

Obtain final gathering according to following steps and close C_final：For the product complete or collected works G={ G with complete lifecycle₁, G₂,…G_v}：

Step a：A product G is chosen from described product complete or collected works G_i, as first cluster C₁={ G_iIn product；

Step b：By G_iRemove from product complete or collected works, take out and in product complete or collected works G, remove G_iSurplus products G_others=G/G_iIn Arbitrary product G_j, for G_j∈G_othersIf, HMSim_i,j>=ε, then by product G_jIt is incorporated in cluster, C₁=C₁∪{G_j}；

Step c：Repeat step b, until no longer have new product to be merged in cluster C₁In, thus obtaining a complete cluster C₁；

Step d：By cluster C₁In included product remove, obtain the product complete or collected works G=G/C updating₁, for the product after updating Product complete or collected works, repeat step a to step c, until producing new cluster again, thus obtain all of cluster C={ C₁,C₂,…C_W, Wherein, W≤V；

Step e：Another threshold value η is set, and wherein, η is integer, and η >=2, and included product number in all clusters is less than η Being combined of cluster obtain C_small={ C₁,C₂,…,C_S, then cast out C_smallCluster in set, obtains final gathering and closes C_final=C/C_small.

14. devices according to claim 9 are it is characterised in that described determining module is additionally operable to：

Determine each life cycle mould in described specified class product life cycle set of modes according to following steps A1 and step B1 The interval position of the life stage of formula：

Step A1：The life cycle of product is corresponding life by the division methods according to default Various Phases of Their Life Cycle In the stage, described life stage includes：Introduction period, growth stage, maturity period, decline phase and end period；The concrete steps of step A1 Including：

If the sales volume curve of the product after product original sales volume curve being fitted using Bass model is：

Y=f (t), t=1,2 ..., T

First derivative is asked to the product sales volume curve after described matching, obtains first derivative curve：

△ f (t)=f (t+1)-f (t), t=1,2 ... T-1

Second dervative is asked to the product sales volume curve after described matching, obtains Second derivative curves：

△²F (t)=△ f (t+1)-△ f (t), t=1,2 ... T-2

Wherein, t represents the time, and T represents the maximum of time；

By corresponding to maximum selling value on the Second derivative curves on the maximum selling value left side of the product sales volume curve after matching Time point as the introduction period end point, simultaneously as the starting point of growth stage,

By the maximum on the product sales volume curve after the time point corresponding to the maximum selling value on first derivative curve and matching The midpoint between time point corresponding to sales volume value as the end point of growth stage, simultaneously as the starting point in maturity period,

By the minimum on the time point corresponding to the maximum selling value on the product sales volume curve after matching and first derivative curve The midpoint between time point corresponding to sales volume value as the end point in maturity period, simultaneously as the starting point of decline phase,

By corresponding to maximum selling value on the Second derivative curves on the right of the maximum selling value of the product sales volume curve after matching Time point as the decline phase end point, simultaneously as the starting point of end period；

Step B1：By each lifetime value in described specified class product life cycle pattern according to the said goods Life Cycle The division methods in phase in each stage determine each stage of the life cycle of product, so that it is determined that the life of each lifetime value The interval position in stage.

15. devices according to claim 9 are it is characterised in that described computing module is additionally operable to：

Assume that in described new product sales volume curve and the lifetime value set of described specified class product, arbitrary lifetime value institute is right The time series answered is respectively：With Wherein, n >=m, represents sequences y respectively⁽¹⁾And y⁽²⁾Time point；x_n,x_mIt is respectively sequences y⁽¹⁾And y⁽²⁾Product when each moment The sales volume value of product；

One new sequence is built according to the time series more than the comprised moment：

Wherein NA represents null value；

By time series y⁽²⁾In time series y^baseOn move, will time series y⁽²⁾The 1st point respectively with time sequence Row y^baseOn i-th point alignment, wherein i=1,2 ..., m+n-1；When time sequences y⁽²⁾The 1st point and time series y^base I-th point of alignment when, definition similarity now is：

${\mathrm{Sim}}_i=\left\{\begin{array}{cc}\frac{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}(y_{m+j-1}^{base}+1)\&CenterDot;(y_{m-j+1}^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}{(y_{m+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{i}{\mathrm{\&Sigma;}}}{(y_{m-j+1}^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2i}{m+n},& 0<i\&le;m-1\\ \frac{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}(y_{i+j-1}^{base}+1)\&CenterDot;(y_j^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{(y_{i+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}{(y_j^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2m}{m+n},& m-1<i\&le;n-1\\ \frac{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}(y_{i+j-1}^{base}+1)\&CenterDot;(y_j^{\left(2\right)}+1)}{\sqrt{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}{(y_{i+j-1}^{base}+1)}^2}\&CenterDot;\sqrt{\underset{j=1}{\overset{n+m-1-i}{\mathrm{\&Sigma;}}}{(y_j^{\left(2\right)}+1)}^2}}\&CenterDot;\mathrm{\&lambda;}\frac{2(n+m-1-i)}{m+n},& n-1<i\&le;n+m-1\end{array}\right.$

Wherein, λ is penalty coefficient；

Calculate time series y⁽²⁾With time series y in moving process^baseOn i-th point alignment when sim_i, thus obtaining The similarity of lifetime value described in the lifetime value set with described specified class product for the described new product：

HMSim=max { Sim₁,Sim₂,...,Sim_n+m-1}；

For the remaining lifetime value in the lifetime value set of described specified class product, repeat above step with Obtain each lifetime value in the sales volume curve of described new product and the lifetime value set of described specified class product Similarity, thus obtaining similarity set.

16. devices according to claim 9 are it is characterised in that described identification module is additionally operable to：

For the moment t=n of new product, the described new product sales volume curve according to record and the life cycle mould of described new product The position of the relative movement of formula determines the position of two recorded curve relative movements to obtain this moment product new product pin Sell the position t of the bass diffusion models curve of the lifetime value that curve corresponds to described new product_bass=m, this position institute Corresponding life cycle phase is the life-cycle stages of this new product；

For this new product life cycle future time instance t=n+i, i=1,2 ..., by the Bath of this new product lifetime value The position t of diffusion model curve_bass=m+i, i=1,2 ... the life cycle phase residing for the moment is as this new product Life Cycle The predicted value of phase future time instance.