Advances n Management & Appled Economcs, vol., no., 0, - ISS: 79-7 (prnt verson), 79-7 (onlne) Internatonal Scentfc Press, 0 Product Innovaton Rsk Management based on Bayesan Decson Theory Yngchun Guo Abstract Innovaton s an nexhaustble force for the prosperty of one naton, and also the lfe source of enterprses. Product nnovaton s an mportant aspect of nnovaton. However, the product nnovaton actvtes has hgh-rsk characterstcs. Enterprses have to perform scentfc and effectve product nnovaton rsk management. Based on a general ntroducton of Bayestan Decson Theory prncple, the author studed the practces of product nnovaton n enterprses. The paper dscussed how to use Bayesan Decson Theory to acheve quanttatve nnovaton-rsk management n product nnovaton: based on the descrpton of three elements for product nnovaton rsk management, the author dscussed the process of bayesan rsk decson-makng n product nnovaton. Thus to provdng references for scentfc decson of nnovaton actvtes n enterprses. College of Mathematcs & Computer Seence, Hebe Unversty, Chna. e-mal: guoyc@hbu.cn Artcle Info: Receved : December, 0. Revsed : January, 0 Publshed onlne : February 8, 0
Product Innovaton Rsk Management based on Bayesan Decson Theory JEL classfcaton numbers: O Keywords: Product Innovaton, Innovaton Rsk, Rsk Management, Bayesan Decson Theory Introducton Innovaton s the soul of a naton s progress, an nexhaustble force for the prosperty of a naton, and the lfe source of enterprses []. Wthout nnovatons, enterprses would not be able to upgrade the producton structure. Wth weakenng compettveness, enterprses wll de. However, nnovaton s a double-edged sword, wth characterstcs of hgh potentals, hgh nputs, hgh returns, and hgh rsks. Partcularly, hgh rsks from technologes, market, and management frustrate or even kll many nnovaton actvtes, whch may even threaten the healthy development of human socety. Therefore, to manage the nnovaton rsks s sgnfcant. Currently, most researches on nnovatons are about methods and modes that promote enterprses to develop ndependent nnovatons, seldom focus on nnovaton rsks. [] bult an applcaton framework for hgh-tech enterprses mplementng overall rsk management. [] proposed a syntheszed rsk management mode for enterprses coorperatve nnovatons based on the meta-synthess method. [] put forward the rsk management strategy n the process of technologcal nnovaton use to acheve effectve rsk preventon. All these lteratures were qualtatve studes on dfferent stages of rsk management. In the theoretcal feld, there are quanttatve researches on nnovaton rsk management. [] proposed the synthess evaluaton method and appled t to the rsk evaluaton of enterprses technologcal nnovaton. [6] bult a rsk pre-warnng system for enterprses technologcal nnovaton proects. [7] proposed a producton nnovaton program drven by market or customer data.
Yngchun Guo These researches promoted the scentfc decson of technologcal rsk management, but the applcaton s unsatsfyng. On one hand, these methods are too complcated to use n enterprses. On the other hand, most quanttatve studes focus on the rsk evaluaton, but seldom on rsk decson. Rsk decson-makng s to make decson accordng to ncomplete nformaton. Accordng to the obectve of rsk management, wth bass of rsk dentfcaton and rsk evaluaton, make reasonable choce and combnaton of dfferent rsk management methods, and offer a specfc program for rsk management. Faced hgh rsks from technologes, market, and management, enterprse managers should master the scentfc and feasble rsk decson-makng method, managng nnovaton rsks effectvely. Bayesan approach s a powerful tool for rsk decson-makng [8]. Due to ts convenence and easness, ths approach s applyng n many felds. [9] used the Bayesan Decson technology to support the new product development management. [0] appled the Bayesan network method to the rsk evaluaton n new product R & D. [] proposed a Bayesan soluton for enterprses predctng the strategc marketng management decson. [] bult a Bayesan model to acheve dynamc knowledge update, n order to deal wth the supply uncertantes and rsks. Ths paper s to explore the effectve quanttatve rsk decson-makng method, n order to help enterprse managers to acheve effectve nnovaton rsk management. An Introducton of Bayesan Decson Theory Rsk decson-makng decson runs through the whole rsk management process. By analyzng rsks and losses scentfcally, t can help to choose the reasonable rsk management technques and methods and fnally get the most satsfyng soluton from several optons. Every rsk decson-makng ncludes
6 Product Innovaton Rsk Management based on Bayesan Decson Theory three elements: the state group conssted of dfferent natural status, the acton group conssted of a set of actons taken by decson makers, and the descrpton of utlty or losses from dfferent combnatons of statues and actons. From the three elements, we can get dfferent rsk condtons. Once the decson maker makes a decson wth uncertan result, t means certan rsk. The rsk decson-makng needs to get changeable market nformaton by ncreasng nputs. Based on masterng varous natural condtons n tme, use the collected nformaton reasonably, and select the decson scentfcally, reducng rsks, and mprovng economc and socal benefts. In rsk decson-makng, the accuracy of estmaton of natural condtons can drectly affect the expected returns. In order to make better decson, t needs to update the nformaton n tme. After gettng new nformaton, we can revse the orgnal estmated probablty of emergence of certan natural condton, and use the revsed probablty dstrbuton to make new decson. Because the probablty correcton s based on the Bayesan Theorem n probablty theory, ths decson s called Bayesan Decson. Product Innovaton Rsk Management Cases. Three Elements for Innovaton Rsk Management Decson.. The Set of atural States The comprehensve evaluaton on nnovaton actvty s,,, m. For nstance, stands for best, stands for better,, and m stands for worst. Experts gve the predcton posteror probablty of each state P,,..., m (See Table )..
Yngchun Guo 7 Table : Utlty State&probablty : P u P u : m Utlty d d dn u un u un P m um um umn : Program.. The Set of Actons The acton toward nnovaton actvty s D d d d,,, n. Here d stands for hgh nvestment, such as more nvestment n R & D, new producton equpment, and new product. d stands for medum nvestment, such as medum nvestment n R & D, and changes of product functons. d stands for low nvestment, such as changes of producton technques, and better product qualty. d stands for no nvestment n nnovaton, such as only changes n packages or more advertsements... The Matrx of Descrptons of Utlty or Losses U u. Here, u -00,00 s the economc utlty that can be mn evaluated by money, or the utlty functon evaluated by non-monetary factors. Here, we suggest the second meanng, because nnovaton actvtes can not only generate economc benefts, but also socal benefts, so as to brng ntangble assets and long-term nterests for enterprses. Here, the utlty functon can be measured by the satsfacton degree, such as enterprses satsfacton degree, customers satsfacton degree, expert scorng, and other comprehensve scores.
8 Product Innovaton Rsk Management based on Bayesan Decson Theory. Descrpton of Product Innovaton Rsk Suppose an enterprse starts a new product R & D. There are fve states of comprehensve evaluatons on economc utlty and socal benefts,,,,. Here, stands for best, stands for better, stands for medum, stands for worse, and stands for worst. Accordng to the data analyss of the market survey and the expert predcton, the probablty dstrbuton of each state s P =0., P =0., P =0., P P. The enterprse has four optons D d, d, d, d =0.0 =0.,. d stands for hgh nvestment, d stands for medum nvestment, d stands for low nvestment, and d stands for no nvestment. The utlty of four optons under dfferent states s n Table. Table : The expected utlty of nvestment State&probablty : : : : : Utlty Program d d d d P =0. u =00 u =70 u =60 u =-80 P =0. u =70 u =80 u =70 u =-60 P =0. u =0 u =60 u =80 u =-0 P =0. u = -0 u =0 u =0 u =-0 P =0.0 u =-00 u =-80 u =-0 u =0 Data descrpton: the expected utlty declnes along wth the dmnshng prospect of market state. For nstance: u : under the hgh nvestment and best market condtons, the economc utlty and socal benefts reach the hghest. The expected utltyu =00; u : under the hgh nvestment and better market condtons, the economc utlty and socal
Yngchun Guo 9 benefts are hgh. The expected utltyu =70; u : under the hgh nvestment and ordnary market condtons, the economc utlty and socal benefts are medum. The expected utlty su =0; u : under the hgh nvestment and worse market condtons, the economc utlty and socal benefts are worse. The expected utlty s u =-0. u : under the hgh nvestment and worst market condtons, the enterprse suffers from serous losses. The expected utlty su =-00. Here, please focus on the last lne. If the enterprse takes the no nvestment strategy, the expected utlty wll be negatve. For nstance, u : the enterprse does not nvest, though the market condtons are good. It wll make the enterprse lose potental economc utlty and socal benefts. The expected utltyu =-80; u : the enterprse does not make nnovaton nvestment and the market condtons are bad. Then, there s no economc beneft or socal beneft. The expected utltyu =0.. The Bayesan Rsk Decson-Makng Process.. Pror Analyss Accordng to the probablty of natural state and the expected utlty (see Table ), by followng the law of expectaton, calculate the expected utlty of each program. E d P u,,...,. Accordngly, the optmal expectaton for the optmal program s max E d E d EMU. For nstance, Ed =0.*00+0.*70+0.*0+0.*(-0)+0.0*(-00)=0; smlarly, Ed =., Ed =8., Ed =-. Then, the optmal decson and the optmal expected utlty s EMU = E d =8.. It means that the enterprse can take the low-nvestment strategy f only wth the pror nformaton. k
0 Product Innovaton Rsk Management based on Bayesan Decson Theory.. Predcton Posteror Analyss In predcton posteror analyss, estmate the value of complete nformaton frstly. As the predcton of complete nformaton s n the state, t becomes the decson-makng under certanty. Apparently, the optmal program s k k max u. Then, wth complete nformaton, the maxmum expected utlty from decson-makng s: k k k EUPI P max u =0.*00+0.*80+0.*80+0.*0+0.0*0=7.. Therefore, the value of complete nformaton EVPI EUPI EMU =7.-8. =. It means the value of complete nformaton s equal to unts of utlty... Posteror Analyss () Supplement new nformaton. Accordng to the market condtons, nvestgate, explore, and consult the fve states (excellent), (better), (medum), (worse), and (worst), and predct whch one wll appear. Meanwhle, get the condtonal probablty P, whch s the probablty of predctng the emergence of when the natural state actually appears. (See Table ). Lkelhood rato : : : : : Table : The lkelhood rato P P =0. 0. 0. 0. 0. 0.0 P =0. 0. 0. 0. 0.0 0.0 P =0. 0. 0. 0. 0. 0.0 P =0. 0.0 0. 0. 0. 0. P =0.0 0.0 0. 0. 0. 0.
Yngchun Guo () Revse the probablty. Based on the pror probablty P (=,,,) and the condtonal probablty probablty dstrbuton of P (=,,,; =,,,), calculate the : P P P. For nstance, P 0.*0.+0.*0.+0.*0.+0.*0.0+0.0*0.0=0.. Smlarly, P 0.07, P 0.7, P 0., and P 0.08. Use the Bayesan formula to calculate the revsed probablty of, namely the posteror probablty (see Table ): P P P P, (=,,,; =,,,). Posteror probablty Table : The posteror probablty P 0.76 0.80 0.09 0.07 0.09 0.0 0.60 0.0 0.07 0.06 0. 0. 0.00 0. 0.00 0.90 0.90 0.9 0.89 0.06 0.0 0.00 0.0 0.87 0. () Posteror decson. Suppose the supplement nformaton predcts the appearance of state k k. Use the posteror revsed probablty dstrbuton P (=,,,) to calculate the expected utlty of each program. By followng the law of expectaton, make the decson. Then, E d P u, (=,,,, k=,,,). k k For nstance, f the market survey shows that the market condton s, calculate
Product Innovaton Rsk Management based on Bayesan Decson Theory the expected utlty of d k (see Table ). E d =0.76*00+0.8*70+0.09*0+0.07*(-0)-0.09*00=77.. Smlarly, there s Ed 68.9, E d =6., E d -6.8. Here, as the market condton s better, the enterprse can take the stategy d. The maxmum expected utlty s Ed 77.. Smlarly, As the market condton s, the maxmum expected utlty s E d 68.7; As the market condton s, the maxmum expected utlty s Ed 6.6; As the market condton s, the maxmum expected utlty s Ed.9; As the market condton s, the maxmum expected utlty s Ed 8.. Table : Posteror expected utlty k The posteror expected utlty E d d d d d 77. 68.9 6. -6.8 6.9 68.7 6.8-6.0 9.9 7.7 6.6-7.68.8 0.6.9 -.8.. 8. -.7 () Calculate the value of supplement nformaton. Accordng to the calculated supplement nformaton, predct the probablty of each status P (=,,,). Calculate the maxmum expected utlty n posteror analyss: EMU *= P E =0.*77.+0.07*68.7+0.7*6.6+0.*.9+0.08*8.=6.
Yngchun Guo Apparently, after gettng the supplement nformaton, the expected utlty rses: EMU * EMU = 6.-8.=.8. The value of supplement nformaton s.8 unt of utlty. Then, compare the value of supplement nformaton and the cost for acqurng the nformaton, and make the rght decson. Concluson The nnovaton rsk management s crtcal for the survval and the development of enterprse. In ths paper, takng the product nnovaton actvty for nstance, the author dscusses the nnovaton rsk management based on Bayesan Rsk Decson-Makng. Here, one pont should be noted partcularly: the repettve applcaton of Bayesan Rsk Decson-Makng can help the enterprse to carry out the dynamc rsk management of nnovaton actvtes and adapt to the changng market condtons, achevng the scentfc management of nnovaton rsks. ACKOWLEDGEMETS. Ths paper s a part of the proect Research on Innovaton Rsk Management Mechansm of Industral Chan of Hebe Soft Scence Program of Chna. References [] S. Cheng, On the constructon of an nnovatve country, Chna Soft Scence,, (009),-. [] F. ao and J. Hao, On the ERM mplementaton framework of hgh-tech enterprses, Scentfc Management Research, 8(), (00), 66-69.
Product Innovaton Rsk Management based on Bayesan Decson Theory [] R. Lu and K. Wang, Research on ntegrated management for enterprse s cooperaton nnovaton, Scence and Technology Management Research, 0, (009), 9-0. [] Y. Mao, Enterprses n the process of technologcal nnovaton rsks and preventon strategy, Scence and Technology Management Research, 7, (00), -. [] Z. Song, S. Wang and Z. Lu, Applcaton of AP-GRAP ntegratng method for rsk evaluaton n enterprses' technologcal nnovaton, Scence of Scence and Management of S. & T,, (00), -8. [6]. L, J. u and J. Yan, Study on the constructon of rsk early warnng system for enterprse technologcal nnovaton proects, Journal of Schuan Unversty (Socal Scence Edton),, (00), 88-9. [7] K. Andrew, Innovaton: a data-drven approach, Internatonal Journal of Producton Economcs, (), (009), 0-8. [8] B. Rchard, A unfed Bayesan Decson Theory, Theory and Decson, 6(), (007), -6. [9] P.V. Jacobus and C.V.W. Cornels, ew product development wth dynamc decson support, Internatonal Journal of Innovaton and Technology Management, 6(), (009), -67. [0] C. Kwa-Sang, T. Da-We, Y. Jan-Bo, etc, Assessng new product development proect rsk by Bayesan network wth a systematc probablty generaton methodology, Expert Systems wth Applcatons, 6(6), (009), 9879-9890. [] L. Paul and G.L. Reynolds, Predctve strategc marketng management decsons n small frms: A possble Bayesan soluton, Management Decson, (6), (007), 08-07. [] C. Mn,. Yusen and W. nle, Managng supply uncertantes through Bayesan nformaton update, IEEE Transactons on Automaton Scence and Engneerng,, (00), -6.