Industrial Management & Data Systems Examining technological innovation of Apple using patent analysis Sunghae Jun Sang Sung Park
Article information:
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
To cite this document: Sunghae Jun Sang Sung Park, (2013),"Examining technological innovation of Apple using patent analysis", Industrial Management & Data Systems, Vol. 113 Iss 6 pp. 890 - 907 Permanent link to this document: http://dx.doi.org/10.1108/IMDS-01-2013-0032 Downloaded on: 04 April 2016, At: 05:54 (PT) References: this document contains references to 30 other documents. To copy this document:
[email protected] The fulltext of this document has been downloaded 3681 times since 2013*
Users who downloaded this article also downloaded: (2014),"Why Apple’s product magic continues to amaze – skills of the world’s #1 value chain integrator", Strategy & Leadership, Vol. 42 Iss 6 pp. 17-28 http://dx.doi.org/10.1108/SL-10-2014-0074 (2014),"Apple was sweeter when Steve Jobs held sway: Company illustrates contrasting attitudes to organizational change", Human Resource Management International Digest, Vol. 22 Iss 4 pp. 25-28 http:// dx.doi.org/10.1108/HRMID-07-2014-0084 (2014),"Financial decision-making in a high-growth company: the case of Apple incorporated", Management Decision, Vol. 52 Iss 9 pp. 1591-1610 http://dx.doi.org/10.1108/MD-10-2013-0557
Access to this document was granted through an Emerald subscription provided by emerald-srm:458072 []
For Authors If you would like to write for this, or any other Emerald publication, then please use our Emerald for Authors service information about how to choose which publication to write for and submission guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com Emerald is a global publisher linking research and practice to the benefit of society. The company manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as providing an extensive range of online products and additional customer resources and services. Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive preservation. *Related content and download information correct at time of download.
The current issue and full text archive of this journal is available at www.emeraldinsight.com/0263-5577.htm
IMDS 113,6
Examining technological innovation of Apple using patent analysis
890
Sunghae Jun
Received 17 January 2013 Revised 20 March 2013 Accepted 22 March 2013
Department of Statistics, Cheongju University, Chungbuk, Korea, and
Sang Sung Park Division of Information Management Engineering, Korea University, Seoul, Korea
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
Abstract Purpose – Apple is a representative company of technological innovation (TI) and management. It has launched new and innovative products since 1977, and many companies and business schools around the world have attempted to learn about the success story of Apple’s innovation. However, most previous research works on Apple’s innovation have been based on qualitative approaches such as experts’ opinions. Such studies offer a subjective point of view. By contrast, in this paper the authors aim to study the TI and forecasting of Apple by analyzing its patent applications, which is an objective approach to examining the innovation of Apple from a technological perspective. Design/methodology/approach – TI is an important issue concerning technology management for companies and governments. To examine Apple’s TI, the authors analyze all applied patents and construct analytical models according to three approaches. First, they build statistical models using the time series regression and multiple linear regression methods to create a technology map. Second, they cluster all Apple’s patents to find its vacant technology domain. Lastly, they use social network analysis to search for technologies central to Apple’s future. Findings – The authors’ study shows the technological trends and relations between Apple’s technologies. This research finds vacant technology areas and central technologies for Apple’s TI. Practical implications – Using statistical and machine learning methods, the authors analyze all Apple’s patents in order to predict the firm’s future technologies. This research contributes to examining the TI of Apple. Therefore, the results of the patent analysis can highlight the technological opportunities for Apple’s TI. Originality/value – Traditional TI models have been based on qualitative methods. Previous investigations of Apple’s TI have also relied on traditional analytical approaches. In this paper, however, the authors develop a quantitative and objective approach for examining Apple’s TI. Keywords Apple Inc, Technological innovation, Vacant and central technologies, Regression, Patent clustering, Social network analysis, Innovation, Regression analysis, Patents, Social networks Paper type Research paper
Industrial Management & Data Systems Vol. 113 No. 6, 2013 pp. 890-907 q Emerald Group Publishing Limited 0263-5577 DOI 10.1108/IMDS-01-2013-0032
1. Introduction Most companies constantly strive for technological innovation (TI; Cesaratto et al., 1991; Sun et al., 2008), which is not only important for creating competitive advantage but also one of the strategies for generating sustainable development (Chen et al., 2007; Sun et al., 2008; Trappey et al., 2012). Indeed, most successful businesses, including Apple, aim for TI (Mann, 2003), and thus many companies and business schools have attempted to learn about the success story of Apple from a management point of view. The results of previous studies have mostly been based on qualitative methods.
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
In other words, they have depended on the knowledge of experts. Thus, we need more objective and verifiable approaches for examining Apple’s TI, because understanding the state of TI in Apple is necessary for formulating R&D policies and creating new products. Previous research on TI has been based on simple visualizations such as graphs and charts (Cesaratto et al., 1991; Sun et al., 2008; Trappey et al., 2012) or descriptive statistics such as frequency, mean, and index values (Chen et al., 2007). Using these approaches, however, limits our understanding of the TI of Apple because such simple analytical methods are restrictive. Therefore, in this study we examine the TI of Apple using patent analysis, because patents represent a major objective outcome of researched and developed technology (Hunt et al., 2007; Roper et al., 2011). Specifically, we use statistical methods and machine learning algorithms as quantitative methods in the presented patent analysis ( Jun and Uhm, 2010). In this paper, we propose three quantitative models for analyzing Apple’s patents. First, we use time series regression and multiple linear regression models (Bowerman et al., 2005) in order to construct a technology map of Apple. Second, we perform cluster analysis (Jun et al., 2012) on the patent documents in order to predict Apple’s future technologies. Lastly, we formulate the technology networks of Apple’s technologies using social network analysis (SNA) methods (Butts, 2008; Jun, 2012). To provide novel knowledge on Apple’s TI, we then combine the results of our proposed models. Therefore, our research is a meaningful and objective approach to understanding Apple’s innovation. In summary, this study analyzes Apple’s patents around the world in order to examine to Apple’s TI. 2. Research background The performances of a considerable number of leading IT companies have recently declined because they failed to adapt to changes over time and insisted on retaining their traditional technologies. It is recognized that there is no permanently strong company in the IT market. To survive in the competitive IT environment, a company has to bring about technological development and innovate constantly. In 2012, Apple and Samsung were growing consistently through new product development. However, these firms also knew that their market shares could collapse at any time if their TI stopped. After the death of Steve Jobs, the market assessment of Apple was relatively poor. Apple was accused by Samsung of the infringement of its patent and its net profits decreased. Now was the time to change the technological evolution of Apple. 3. Patent analysis for Apple Traditional patent analysis approaches have mainly used citation information (Fattori et al., 2003; Indukuri et al., 2008; Yoon and Park, 2007; Tseng et al., 2005, 2007). Recently, however, some research works have analyzed patent titles and abstracts using text-mining techniques (Jun et al., 2010, 2012). In this paper, we use International Patent Classification (IPC) codes as well as the title and abstract of patent as the input data for the presented patent analysis. IPC codes provide a hierarchical structure of patent classifications based on technological utility (IPC, 2012). Therefore, analyzing the IPC code data extracted from Apple’s applied patents can allow us to understand its technological change and make forecasts. Our analysis consists of three distinct approaches, namely statistical analysis, patent clustering, and SNA. Figure 1 shows our model for examining Apple’s TI.
Technological innovation of Apple 891
IMDS 113,6
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
892
Figure 1. Proposed analytical process for examining Apple’s patents
We retrieve documents on the applied patents of Apple from global patent databases (USPTO, 2012). These documents provide diverse information such as patent title, abstract, IPC code, application number and date, issue number and date, drawings, claims, and so forth (Hunt et al., 2007). In this paper, text mining was used as a preprocessing technique to extract the IPC codes and construct patent-term matrix from the retrieved patent documents. In our text mining processing, we make text corpus based on a collection of titles and abstracts in the patent documents. Next, we get a structured text data by parsing from the text corpus, and we extract IPC codes from the patent documents. So, we construct a patent-term-IPC code matrix as shown in Figure 2. The row and column of this matrix represent patent and term or IPC code, respectively. Each element is a frequency of term or IPC code in each patent. For example, freqP1T2 is the frequency of term2 in patent1. In this paper, we build analytical models for examining Apple’s TI using our structured patent-term-IPC code matrix by text mining technique. As shown in Figure 1, our proposed model consists of three approaches. First, we perform a statistical analysis in order to summarize Apple’s patent data and construct a time series regression model to ascertain its technological trends over time. This analysis allows us to understand Apple’s speed of technological development and to find its so-called target technologies as well as those technologies that influence them.
Figure 2. Patent-term-IPC code matrix
Second, we cluster the patent data. By using the Silhouette width approach (Rousseeuw, 1987; Everitt et al., 2001), we determine the optimal number of clusters (K) and use the K for the K-mean clustering algorithm (Han and Kamber, 2005; Hastie et al., 2001). Patent clustering makes it possible to predict vacant areas of Apple’s TI. Further, the result of this patent clustering is used as the input data for the SNA approach. Lastly, we apply SNA techniques (Cho et al., 2012; Sternitzke et al., 2008) in order to ascertain the elaborate relations between Apple’s technologies. To this end, we construct advanced SNA graphs to draw associations between Apple’s representative technologies and its technological groups. Next, we explain these three approaches in detail.
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
3.1 Statistical analysis To understand Apple’s technologies, we apply a time series regression model (Bowerman et al., 2005) to assess the number of applied patents by year as follows: y t ¼ b0 þ b1 t þ 1 t
ð1Þ
where yt is the number of applied patents in year t, b0 and b1 are the intercept and slope of the trend in year t, respectively, and 1t is the error term in year t. This is a trend model of Apple’s developed technologies. The slope value of this model shows the development speed of the firm’s target technologies. For example, the trend of the IPC code H04M by year is represented as follows: IPC ¼ b0 þ b1 year
ð2Þ
The larger the value of b1, the more the trend increases. In this paper, we select the important technologies according to b1 of Apple’s IPC codes and measure the model utility using the adjusted multiple coefficient of determination (adj. R 2 ) (Bowerman et al., 2005; Han and Kamber, 2005). If adj. R 2 is 1, then the model utility is perfect. After selecting the meaningful IPC codes using the time series regression model, we use a multiple linear regression model (Bowerman et al., 2005) to build a technology map of Apple as follows: IPC T ¼ b0 þ b1 IPC 1 þ b2 IPC 2 þ · · · þ bk IPC k þ 1
ð3Þ
where IPCT is the target IPC code (i.e. the dependent variable in the multiple regression) and IPCi is the ith input IPC code (independent variable). The technological development of the target IPC code is affected by the development of the technologies based on input IPC codes. b0 is the intercept and the mean value of IPCT when all input IPC codes equal 0. bi is the slope and the increased change in the mean value of IPCT associated with a unit increase in IPCi. If bi is negative, the mean value of IPCT decreases as IPCi increases. Moreover, this research carries out t-tests (Ross, 1996) to assess the significance of the parameter bi. The null and alternative hypotheses are as follows: H 0 : bi ¼ 0
vs
H 1 : bi – 0
ð4Þ
The null hypothesis H0 states that the parameter of IPCi is not significant, while the alternative hypothesis H1 represents the statistical significance of the parameter. Using the p-value (probability value) (Ross, 1996) derived from the results of the t-tests, we can support H0 or H1. If the value is smaller than 0.05 (i.e. 95 percent confidence interval), we reject H0 and support H1. This means that bi is statistically significant.
Technological innovation of Apple 893
IMDS 113,6
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
894
Therefore, we can explain the technological development of IPCT by the technological development of IPCi. Using the significant input IPC codes, we thus construct a technology map for the target IPC code. Figure 3 proposes an example of the structure of our technology map. We find that the technological development of the target IPC code (IPCT) is affected by the technological developments of IPCi and IPCj. Further, IPCi and IPCj are influenced by IPCk and IPCp, respectively. For example, the technological development of IPCi affects the target technology IPCT by the weight of bi, and IPCk influences the development of IPCT technology by the value of bk*bi. Therefore, we can construct the technology map of Apple using the result of our statistical analysis of Apple’s patents. 3.2 Patent clustering In the next step, we carry out a cluster analysis of the patent data to understand the technological distribution and ascertain vacant technology areas. As noted earlier, we preprocess the retrieved patent documents using text mining (Tseng et al., 2005, 2007). Our clustering uses the textual data such as patent title and abstract from the patent documents. Figure 4 shows the patent clustering process. Before data clustering, we have to determine the optimal number of clusters (Han and Kamber, 2005). Choosing the number of clusters is crucial for clustering in order to avoid distorted information (Everitt et al., 2001; Han and Kamber, 2005). Generally, this choice depends on the investigator’s experience and thus
Figure 3. Proposed structure of our technology map
Figure 4. Proposed patent clustering process
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
it is a subjective approach. However, many objective methods for choosing the number of clusters have been suggested (Everitt et al., 2001). In this paper, we consider the Silhouette width to be a measure of the number of clusters (Rousseeuw, 1987). In previous works on patent clustering, this measure has shown good performance (Jun and Uhm, 2010; Jun et al., 2012). This method measures the standardized difference between objects by taking the average distance of an object from all others in its cluster and in the nearest cluster (Everitt et al., 2001). Based on this approach, we choose the optimal number of clusters when the Silhouette width is the largest. Then, we perform patent data clustering using the K-means algorithm, where K is the number of clusters chosen by the Silhouette measure. This algorithm starts from an initial K clusters and uses the Euclidean distance between objects and clusters during given iterations. When the change from objects to clusters is hardly shown, the clustering iteration is stopped. To ascertain the vacant or undeveloped technologies of Apple, we thus extract the top ten IPC codes from K clusters in order to predict the firm’s vacant technology.
Technological innovation of Apple 895
3.3 Social network analysis SNA is a method for analyzing the relations between objects in a social system (Butts, 2008; Cho et al., 2012). By using SNA ( Jun, 2012), we can take a network structure of IPC codes and find the central IPC code in the social technology system. In addition, we can understand the mutual relations between IPC codes. Figure 5 shows our SNA process for analyzing Apple’s patents. From the result of our patent clustering analysis, we extract IPC codes in order to construct an IPC code-cluster matrix (ICM). Our ICM consists of clusters and their included IPC codes. We use this ICM to make five SNA graphs: an SNA graph of IPC codes and clusters, a cluster-SNA graph, an IPC code-SNA graph, a cluster-SNA graph based on mutual information, and an IPC code-SNA graph based on mutual information. These graphs illustrate the technological relations between Apple’s IPC codes. 4. Results To verify the performance of the proposed model for examining Apple’s TI, we used the documents of all the patents applied for by Apple. We retrieved patent data in July 2012
Figure 5. Proposed SNA process
IMDS 113,6
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
896
from the US patent databases (USPTO, 2012). Also, we got global patents of Apple via a paid service (WIPSON, 2012). Specifically, we searched for patent documents assigned to Apple because we focused on the technologies from the inside of Apple for examining Apple’s TI. In this paper, we constructed our patent data by combining the searched results from USPTO and WIPSON, and removing the duplicated and non-validated patents. For our experiment, the total number of Apple patents was 8,119 from the period 1977 to 2012. They covered all patents of Apple including applied and utility patents. Figure 6 shows the number of applied patents by year. In Figure 6, we omitted the patents of 2011 and 2012 because our data retrieval did not go through until 2012, while many 2011 patents have not yet been registered in patent databases. The first peak in Apple patents was in 1995 (n ¼ 346). In the early to mid-1990s, Apple researched and developed new computer-related items such as the “Power Macintosh” based on the PowerPC processor of IBM. Since these “glory days,” Apple has faced some difficulties. New challengers such as IBM and Microsoft have threatened its business by providing competitive products such as cheap hardware and imitated graphic user interface (GUI) systems. These were Apple’s anchor products. Thus, the competitive position of Apple weakened in the market, which caused its R&D activities to shrink (see the decrease in patents in the late 1990s). In the early 2000s, Apple became interested in user experience (UX) design, which aims to construct models for explaining all aspects of human-system experiences including products and services. In 2001, the firm introduced the iPod (a portable media player) into the market, with iTunes following thereafter. This was a signal of Apple’s revival. Since then, its number of patents has increased dramatically (including the iPhone and Apple TV). To ascertain the share of technological patents to total patents, we tabulated the representative IPC codes of patents. The representative IPC code is the main IPC code in each patent. We removed patents lacking representative IPC codes. 1,200
1,000
800
600
400
200
0 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Figure 6. Number of Apple patents by year
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
Table I shows the technological dispersion of Apple patents. We found that all Apple patents were based on 164 IPC codes, while the IPC code of G06F accounted for 46 percent of all Apple patents. G06F is represented by the technology domain of “electric digital data processing”. In other words, many of Apple’s technologies depend on this technology domain. The second largest IPC code was G09G which is “arrangements or circuits for the control of indicating devices using static means to present variable information”, while the third was H04N, or “pictorial communication, e.g. television”. In conclusion, most Apple technologies focus on the domain of “digital information control and television.” Next, we analyze the number of patents that have G06F, G09G, and H04N codes. Figure 7 shows that the trend of G06F codes by year is similar to that for the total number of Apple patents. The trajectory of G06F technology increased rapidly from 2004 to 2007 before slowing. Figure 8 shows the numbers of the second to sixth most frequent IPC codes, namely G09G, H04N, G06K, H05K, and H04L. Again, we found that the time series data of these IPC G06F G09G H04N G06K H05K H04L G06T H01R H04B G10L H04M G06Q H04R G11C H03M H04W H01Q H01H G08B H02J H03K G01R G10H H04J G11B G01C H03G G02F H01M H04Q F21V H01L H01J
Frequency
IPC
Frequency
IPC
Frequency
IPC
Frequency
IPC
Frequency
3,337 542 365 242 233 196 195 189 138 127 99 91 76 75 67 66 60 57 56 55 48 42 42 41 38 37 33 31 30 26 24 21 20
G03B H05B B65D H01F G05B H04H G05F H02M A61B B32B G01J A47B G08C G02B G09B H01B A63B B29C B41J H03B G01P H03L A63F E05C H03F B23P H02H H03D G01L G01N G01S G09F H04K
19 19 18 18 17 16 15 15 14 14 14 13 13 12 12 12 11 11 11 11 10 10 8 8 8 7 7 7 6 6 6 6 6
A47G E05D G01D G05D H02G B24B F16M G06N A61M B41B E05F F16L G01B G01K G04B G04G G10K H01P H04S A01H A01K A41D A45C A47F A63H B44C B65B B65H C23C C25D E04B E05B E06B
5 5 5 5 5 4 4 4 3 3 3 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2
F28D G01F G01H G06E G06G G07F H02B H02K H02N H04C A24F A41J A44B A61K A61N B06F B08B B21C B21D B22D B23K B25B B28B B29D B41F B42D C03C C06F C06K C07D C08K C11D C12N
2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
C12P C22B C23F E04C E04F E05G E21B F01D F04D F16H F21S F25B F26B F28F G00F G03C G03F G03G G04F G06H G06I G07G G08F G60F H01C H01T H03C H03H H05F H05H H10R J06K Total
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 7,252
Technological innovation of Apple 897
Table I. Representative IPC code frequencies
IMDS 113,6
450 400 350 300
898
250 200 150 100
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
0
90 80 70 60 G09G (2nd)
50
H04N (3rd)
40
G06K (4th) H05K (5th)
30
H04L (6th)
20
Figure 8. Numbers of the second to sixth most frequent IPC codes by year
10 0 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
50
Figure 7. Number of G06F IPC codes by year
codes had a cyclic linear trend. In terms of the next most frequent codes, Figure 9 confirms that G06T, or “image data processing or generation,” has increased recently. This implies that Apple has developed many image processing technologies in recent years. According to Table II, the top ranked IPC codes of Apple showed few changes throughout the 2000s. This finding suggests that the firm’s major technologies have not changed in the past decade. Next, we constructed a technology map using time series regression and multiple linear regression models. In this experiment, we used the following 11 top ranked IPC codes as follows: G06F (1st), G09G (2nd), H04N (3rd), G06K (4th), H05K (5th), H04L (6th), G06T (7th), H01R (8th), H04B (9th), G10L (10th), and H04M (11th). First, we used the time series regression model to select the IPC code of the target technology. In this regression, the independent variable was time (year). Then, we constructed 11 regression models according to the 11 top ranked IPC codes. Figure 10 shows the slope (b1) values of these 11 IPC codes.
Technological innovation of Apple
35 30 25 G06T (7th)
20
899
H01R (8th)
15
H04B (9th) G10L (10th)
10
H04M (11th)
Figure 9. Numbers of the seventh to eleventh most frequent IPC codes by year
0 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
5
Year
Top ranked IPC code
1st
2nd
3rd
2010 2009 2008
4th
5th
6th
7th
8th
G06K
G06T
H05K
H01R
H04L
G06T, H04L
9th
H04B G10L
G11C H03M
H01R
G06K, H05K
G06K
H05K
G06T
H04L
H01R
H04M
G10L
H04B
H05K
G06T
H01R
G06K
G10L H04M
H04N 2007
H04L
G09G 2006 2005
G06Q, H05K G06K
G06F
H05K
2004
H01R
2003
G06T
G06K, H04B, H04N H04Q
H04B H04R
G01C, G06K, G06T
H01R, H04B
H05K
2001
H04N H05K
2000
G09G
H04L
G06T
G09G, H04L H04N
G06K
G11C
G10H, G10L, H01R, H04W H05K
H04M
G06T
H04L
H05K
H01H
H04N 2002
10th
G10L, H01R, H02J
G11C, H04L
G06Q, G10L, H01R, H03M
G01R, G06K, H05K
G06Q, G06T, G08B G10H, H01L, H01R
Table II. Top ten IPC codes in the 2000s
IMDS 113,6
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
900
Figure 10. Slope (b1) values of the time series regression model by IPC code
The slope value represents the influence of each IPC code. Because G06F has the largest b1 values, we selected this IPC code as Apple’s target technology. Figure 11 shows the adjusted R 2 values of the regression model by IPC code. Again, the value of G06F was the largest. Therefore, we concluded that G06F was the representative IPC code of all Apple’s technologies and used it as the dependent variable. Using the results from the time series regression model, we then built a technology map of Apple using the multiple linear regression model. Table III shows the first results of the multiple linear regression model. Table III shows which codes influenced the dependent variable of G06F. Those IPC codes that have a p-value less than 0.05 are considered to affect the technology domain of G06F. The IPC codes of G09G, H04N, and G10L had a statistically significant influence on the development of G06F. The connecting weights to G06F were the
Figure 11. Adjusted R 2 values of the time series regression model by IPC code
regression parameters (b1). These parameters represent the increase in G06F associated with a one-unit rise in G09G, H04N, and G10L. Of these codes, G10L has the largest influence on G06F because the parameter of G10L is the largest. To complete Apple’s technology map, we thus performed a second multiple linear regression (Table IV). In this case, we found three dependent variables, namely G09G, H04N, and G10L. These dependent variables were thus used as independent variables in the first case.
Dependent code
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
G06F
Dependent code G09G
H04N
G10L
Independent code
Parameter (b1)
p-value
G09G (2nd) H04N (3rd) G06K (4th) H05K (5th) H04L (6th) G06T (7th) H01R (8th) H04B (9th) G10L (10th) H04M (11th)
1.36 4.12 0.61 21.25 0.55 21.47 1.39 20.67 7.81 20.22
0.007 0.001 0.655 0.473 0.645 0.308 0.312 0.679 0.001 0.898
Independent code
Parameter (b1)
p-value
H04N (3rd) G06K (4th) H05K (5th) H04L (6th) G06T (7th) H01R (8th) H04B (9th) G10L (10th) H04M (11th) G06K (4th) H05K (5th) H04L (6th) G06T (7th) H01R (8th) H04B (9th) G10L (10th) H04M (11th) G09G (2nd) H04M (11th) G09G (2nd) H04N (3rd) G06K (4th) H05K (5th) H04L (6th) G06T (7th) H01R (8th) H04B (9th)
1.43 1.42 21.74 0.82 20.80 20.14 0.42 0.06 0.27 20.31 0.91 0.10 0.30 20.13 0.37 0.21 20.41 0.24 0.10 0.00 0.07 20.08 0.14 0.03 0.29 0.11 0.04
0.002 0.011 0.016 0.111 0.200 0.814 0.553 0.945 0.716 0.199 0.001 0.628 0.233 0.585 0.194 0.556 0.162 0.002 0.581 0.945 0.556 0.574 0.420 0.815 0.043 0.425 0.809
Technological innovation of Apple 901
Table III. First results of the multiple linear regression model
Table IV. Second results of the multiple linear regression model
IMDS 113,6
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
902
Figure 12. Apple’s final technology map
Figure 13. Silhouette width by number of clusters
From the first result, we already knew that the IPC codes of G06K, H05K, and H04N affected the development of G09G. Further, we found that H05K and G06T influence H04N and G10L, respectively. From the results of the first and second models, we therefore constructed the final technology map for examining Apple’s TI (Figure 12). From Figure 12, we conclude that Apple’s most fundamental technology is the technology domain of G06F. However, the technological paths of Apple can also be explained by the technology map. For example, G06K affects G09G by a weight of 1.4246, while G09G influences G06F by a weight of 1.3638. Further, G06K affects G06F through G09G by a weight of 1.4246*1.3638. The remaining paths can also be described in this way. Moreover, we can determine the number of clusters with the largest Silhouette value. According to the results shown in Figure 13, the number was five. Using this result, we performed K-means clustering with K ¼ 5. Table V shows the clustering results. Because most patents were in cluster 1, the technology of cluster 1 cannot become the vacant field of Apple’s TI. Cluster 3 also has several patents, whereas clusters 2 and 5 have relatively few patents. Hence, we considered these two clusters to be Apple’s vacant technology areas. Table VI shows the representative IPC codes of these five clusters. The IPC codes of G06F, H04B, and H04N are included in all clusters. This means that most technologies of Apple are based on the technologies of G06F, H04B, and H04N IPC codes.
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
The defined technologies of G06F, H04B, and H04N are “electric digital data processing”, “transmission”, and “pictorial communication, e. g. television”, respectively. We knew most of Apple’s patents were developed by the technologies of these codes. They are hardly conductive to separate out their detailed technologies from the five clusters. So, we did not use them to characterize each cluster. The IPC codes of H04R, G10H, H03G, and H02B are used to separate cluster 5 from the others because these codes are in this cluster exclusively. According to the results of our patent clustering, we can thus define cluster 5 as the vacant technology area for examining Apple’s TI according to the World Intellectual Property Organization (IPC, 2012). The defined technologies include “the communication and processing technologies of electronic music and speech.” Therefore, Apple needs to develop these technologies relatively more than it does the others (Table VI). The final approach in the presented analysis was the construction of SNA graphs. Using the results from the patent clustering, we built different SNA graphs for examining Apple’s TI based on the ICM. The row and column of ICM represent the IPC code and cluster, respectively, and an element of this matrix has binary value. When the value is one, the technological relationship between the IPC code and the cluster is existed. Using the binary nature of the ICM, we can check whether an IPC code occurs in each cluster. In other words, if the value is 1, then the IPC code occurred in the cluster and 0 otherwise. Figure 14 shows which IPC codes are associated with which technology clusters. The IPC codes that connect to clusters relatively more frequently represent meaningful
Technological innovation of Apple 903
Number of patents
%
5,852 351 1,071 676 169
72.08 4.32 13.19 8.33 2.08
Table V. K-means clustering results
Electric digital data processing Speech analysis or synthesis; speech recognition; audio analysis or processing Pictorial communication, e. g. television Electrically-conductive connections; structural associations of a plurality of mutuallyinsulated electrical connecting elements; coupling devices; current collectors H04R Loudspeakers, microphones, gramophone pick-ups or like acoustic electromechanical transducers; deaf-aid sets; public address systems G10H Electrophonic musical instruments; instruments in which the tones are generated by electromechanical means or electronic generators, or in which the tones are synthesized from a data store H04M Telephonic communication H04B Transmission H03G Control of amplification H02B Boards, substations, or switching arrangements for the supply or distribution of electric power
Table VI. Defined technologies of the IPC codes in cluster 5
Cluster Representative IPC codes 1 2 3 4 5
G06F, G09G, H04N, H05K, G06K, H01R, H04L, G06T, H04B, H04M G06F, G06Q, G11B, H04N, H04B, H04H, G09G, H03G, G10L, H01R G06F, G09G, H04N, G06T, G06K, H04L, G10L, H01R, H02J, H04B G06F, H04N, G09G, H04L, H03M, G06K, H04B, G11C, G06T, H04J G06F, G10L, H04N, H01R, H04R, G10H, H04M, H04B, H03G, H02B
IPC code
Defined technologies
G06F G10L H04N H01R
IMDS 113,6
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
904
Figure 14. SNA graph of IPC codes and clusters
technologies for Apple. We can see that the IPC codes of G06F, G06T, G09G, H01R, H04H, H04L, and H04N are connected to most clusters. Therefore, these are Apple’s meaningful technologies. Next, we formulated an SNA graph between clusters to understand the central technology of Apple (Figure 15). Clusters 1 and 3 are defined as the central technologies of Apple because they have four connecting paths (the largest number). Similar to the technology that defines Apple’s vacant technology (Table VI), we defined the representative technologies of clusters 1 and 3. These clusters have the same IPC codes, namely G06F, G09G, H04N, G06K, H01R, H04L, G06T, and H04B. These are thus common technologies between clusters 1 and 3. However, cluster 1 also has the H05K and H04M codes, while cluster 3 also has G10L and H02J. Therefore, we define the technology of cluster 1 as “the technologies of digital image data controls such as transmission, communication,
Figure 15. Cluster-SNA graph for finding Apple’s central technology
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
recognition, presentation, and arrangement.” In addition, cluster 2 was defined as “the technology of digital audio data controls such as analysis transmission, communication, recognition, and arrangement.” The representative technologies of clusters 1 and 3 are based on “image” and “audio,”, respectively. The integrated technology of clusters 1 and 3 can be defined as “the technologies of digitalized video data handling.” The final SNA graph is the cluster-SNA graph based on mutual information, as shown in Figure 16. The thickness of the connecting lines denotes the association weight between two clusters. Clusters 1, 3, and 4 are interconnected strongly. Figure 15 showed that clusters 1 and 3 were the central technologies of Apple. Further, we can confirm that cluster 5 is weakly connected to clusters 1, 2, and 3 (the technology of cluster 5 was the undeveloped vacant technology according to the result of Table VI). Therefore, we can conclude that our previous results are valid.
Technological innovation of Apple 905
5. Conclusions This paper examined Apple’s applied patents by using three analytical models for the presented patent analysis. They performed independently of each other. First, statistical methods of time series and multiple regressions were applied to create a map of technological relationship. The p-value was used to select the meaningful IPC codes (technologies) and the regression parameters represented the associated weights between technologies. Using this result, we found the G06F code to be the target technology of Apple and the IPC codes of G09G, H04N, G10L, G06K, H05K, and G06T as those that influenced G06F directly and indirectly. Second, Using cluster analysis, we grouped the technologies of Apple to homogeneous clusters. From the clustering results, we selected undeveloped and vacant areas of Apple’s technologies and found that “the communication and processing technologies of electronic music and speech” were the vacant areas for Apple’s TI. Lastly, we built three SNA graphs to ascertain the relations between clusters and IPC codes and found the central technologies as well as checked the undeveloped and vacant technology in Apple uncovered earlier. So, we found technological impact and forecasting of Apple by combining the results of the three approaches. Although traditional TI has been based on creative analyses, we focused on the analysis of the developed technologies of Apple based on its applied patents in order to examine its TI. It is important that the TI of Apple inspires creative designers.
Figure 16. Cluster-SNA graph based on mutual information
IMDS 113,6
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
906
This research contributes to examining the TI of Apple objectively. In our study, we used the class level of IPC code (e.g. G06F, H05K) as the data input for the patent analysis. However, if the group level of IPC code (e.g. G06F 3/100, H05K 3/02) were analyzed, we may ascertain more detailed results. Further, more advanced statistical methods for the patent analysis can be considered to be a novel result for examining Apple’s TI. These advances will be furthered in our future works. In addition, we will compare the TIs of Apple and Samsung by analyzing their patent data. References Bowerman, B.L., O’Connell, R.T. and Koehler, A.B. (2005), Forecasting, Time Series, and Regression: An Applied Approach, Brooks/Cole, Pacific Grove, CA. Butts, C.T. (2008), “Social network analysis with sna”, Journal of Statistical Software, Vol. 24 No. 6, pp. 1-51. Cesaratto, S., Mangano, S. and Sirilli, G. (1991), “The innovative behaviour of Italian firms: a survey on technological innovation and R&D”, Scientometrics, Vol. 21 No. 1, pp. 115-141. Chen, D., Lin, W.C. and Huang, M. (2007), “Using essential patent index and essential technological strength to evaluate industrial technological innovation competitiveness”, Scientometrics, Vol. 71 No. 1, pp. 101-116. Cho, Y., Hwang, J. and Lee, D. (2012), “Identification of effective opinion leaders in the diffusion of technological innovation: a social network approach”, Technological Forecasting and Social Change, Vol. 79, pp. 97-106. Everitt, B.S., Landau, S. and Leese, M. (2001), Cluster Analysis, 4th ed., Edward Arnold, London. Fattori, M., Pedrazzi, G. and Turra, R. (2003), “Text mining applied to patent mapping: a practical business case”, World Patent Information, Vol. 25, pp. 335-342. Han, J. and Kamber, M. (2005), Data Mining: Concepts and Techniques, Morgan Kaufmann, Burlington, MA. Hastie, T., Tibshirani, R. and Friedman, J. (2001), The Elements of Statistical Learning, Data Mining, Inference, and Prediction, Springer, Berlin. Hunt, D., Nguyen, L. and Rodgers, M. (2007), Patent Searching Tools & Techniques, Wiley, New York, NY. Indukuri, K.V., Mirajkar, P. and Sureka, A. (2008), “An algorithm for classifying articles and patent documents using link structure”, Proceedings of International Conference on Web-Age Information Management, pp. 203-210. IPC (2012), International Patent Classification Official Publication, World Intellectual Property Organization (WIPO). Jun, S. (2012), “Central technology forecasting using social network analysis”, Communications in Computer and Information Science, Vol. 340, pp. 1-8. Jun, S. and Uhm, D. (2010), “Patent and statistics, what’s the connection?”, Communications of the Korea Statistical Society, Vol. 17 No. 2, pp. 205-222. Jun, S., Park, S. and Jang, D. (2012), “Technology forecasting using matrix map and patent clustering”, Industrial Management & Data Systems, Vol. 112 No. 5, pp. 786-807. Jun, S., Park, S., Shin, Y., Jang, D. and Chung, H. (2010), “Forecasting vacant technology of patent analysis system using self organizing map and matrix analysis”, Journal of the Korea Contents Association, Vol. 10 No. 2, pp. 462-480. Mann, D.L. (2003), “Better technology forecasting using systemic innovation methods”, Technological Forecasting and Social Change, Vol. 70, pp. 779-795.
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
Roper, A.T., Cunningham, S.W., Porter, A.L., Mason, T.W., Rossini, F.A. and Banks, J. (2011), Forecasting and Management of Technology, Wiley, New York, NY. Ross, S.M. (1996), Introductory Statistics, McGraw Hill, New York, NY. Rousseeuw, P.J. (1987), “Silhouettes: a graphical aid to the interpretation and validation of cluster analysis”, Journal of Computational and Applied Mathematics, Vol. 20, pp. 53-65. Sternitzke, C., Bartkowski, A. and Schramm, R. (2008), “Visualizing patent statistics by means of social network analysis tools”, World Patent Information, Vol. 30, pp. 115-131. Sun, Y., Lu, Y., Wang, T., Ma, H. and He, G. (2008), “Pattern of patent-based environmental technology innovation in China”, Technology Forecasting and Social Change, Vol. 75, pp. 1032-1042. Trappey, A.J.C., Trappey, C.V., Wu, C. and Lin, C. (2012), “A patent quality analysis for innovative technology and product development”, Advanced Engineering Informatics, Vol. 26, pp. 26-34. Tseng, Y., Lin, C. and Lin, Y. (2007), “Text mining techniques for patent analysis”, Information Processing and Management, Vol. 43 No. 5, pp. 1216-1247. Tseng, Y., Juang, D., Wang, Y. and Lin, C. (2005), “Text mining for patent map analysis”, Proceedings of IACIS Pacific Conference, pp. 1109-1116. USPTO (2012), The United States Patent and Trademark Office, available at: www.uspto.gov WIPSON (2012), WIPS Co., Ltd., available at: www.vipson.com Yoon, B. and Park, Y. (2007), “Development of new technology forecasting algorithm: hybrid approach for morphology analysis and conjoint analysis of patent information”, IEEE Transactions on Engineering Management, Vol. 54 No. 3, pp. 588-599. Further reading Camus, C. and Brancaleon, R. (2003), “Intellectual assets management: from patents to knowledge”, World Patent Information, Vol. 25 No. 2, pp. 155-159. McDermott, C.M., Kang, H. and Walsh, S. (2001), “A framework for technology management in services”, IEEE Transactions on Engineering Management, Vol. 48 No. 3, pp. 333-341. Corresponding author Sang Sung Park can be contacted at:
[email protected]
To purchase reprints of this article please e-mail:
[email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints
Technological innovation of Apple 907
Downloaded by BCU BUCURESTI At 05:54 04 April 2016 (PT)
This article has been cited by: 1. Eun Jin Han, So Young Sohn. 2016. Technological convergence in standards for information and communication technologies. Technological Forecasting and Social Change 106, 1-10. [CrossRef] 2. Sangsung Park Graduate School of Management of Technology, Korea University, Seoul, Korea Juhwan Kim Graduate School of Management of Technology, Korea University, Seoul, Korea Hongchul Lee School of Industrial Management Engineering, Korea University, Seoul, Korea Dongsik Jang School of Industrial Management Engineering, Korea University, Seoul, Korea Sunghae Jun Department of Statistics, Cheongju University, Cheongju, Korea . 2016. Methodology of technological evolution for three-dimensional printing. Industrial Management & Data Systems 116:1, 122-146. [Abstract] [Full Text] [PDF] 3. Su Jung Jee, So Young Sohn. 2015. Patent network based conjoint analysis for wearable device. Technological Forecasting and Social Change 101, 338-346. [CrossRef] 4. Farshad Madani, Tugrul Daim, Calvin Weng. 2015. ‘Smart building’ technology network analysis: applying core–periphery structure analysis. International Journal of Management Science and Engineering Management 1-11. [CrossRef] 5. Jong-Min Kim, Sunghae Jun. 2015. Graphical causal inference and copula regression model for apple keywords by text mining. Advanced Engineering Informatics 29, 918-929. [CrossRef] 6. Jungwoo Suh Department of Information and Industrial Engineering, Yonsei University, Seoul, Republic of Korea So Young Sohn Department of Information and Industrial Engineering, Yonsei University, Seoul, Republic of Korea . 2015. Analyzing technological convergence trends in a business ecosystem. Industrial Management & Data Systems 115:4, 718-739. [Abstract] [Full Text] [PDF]