Docstoc

AccSearch: A Specialized Search Engine for Traffic Analysis - PDF

Document Sample
AccSearch: A Specialized Search Engine for Traffic Analysis - PDF Powered By Docstoc
					                                                          (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                    Vol. 8, No. 2, 2010

  AccSearch: A Specialized Search Engine for Traffic
                      Analysis

                    K. Renganathan                                                            B. Amutha
     Computer Science and Engineering Department                            Computer Science and Engineering Department
                   SRM University                                                         SRM University
                        India                                                                  India
               ranganath73@gmail.com                                                  bamutha62@gmail.com


Abstract— AccSearch is a specialized web search engine to              URL result for any queries pertaining to the road accident
provide information about road accidents within Chennai,               details within the Chennai city. It is aimed to provide a high
India and assist the traffic authorities, police, NGOs,                dependency to the user. It covers the entire accident data
lawyers, students and statistical bureaus. The people who              which occurred in the four National Highways around Chennai.
are in need of road accident information for various                   It will provide information about the accident occurred in the
reasons are very much struggling to collect the correct                day and night around the highways. This information is used
information under a single search. Special purpose search              to do the historical collection of data. This traffic search
engines are designed to work on a particular domain                    engine can be later connected to the all purpose search engines
which fill the gap where an all purpose search engine                  to add up the searching power and efficiency.
lacks. As the existing search engines cannot do the traffic
search alone well for several reasons, we have designed a                Markov chain algorithm is used to improve the performance
search algorithm using Markov chain, to provide the                    and speed up. Markov chains are well known for the
search information in a faster manner. The mathematical                performance tuning and prediction.
proof of our modified Markov chain algorithm shows that
the speed and efficiency seems to be better in comparison                 Adding the information with the available information on
with the existing search algorithms. As Markov chain can               the internet is the fruit of this work. By providing some more
be used for prediction purposes, our search engine                     information with the already available information some
concentrates on one particular domain which is traffic                 sectors will be highly benefited. Those include Police, NGOs,
analysis it will result in exact responses to the user queries         Statistical Bureaus and Universities to name a few. It will
and will lead to a greater amount of user satisfaction.                provide a greater benefit to the society.

   Keywords; AccSearch; road traffic; accident;Markov                  A. Literature Survey / Related Works
chain;accident prediction;                                             Sergey Brin and Lawrence Page, “The Anatomy of Large-
                                                                       Scale Hyper textual Web Search Engine” addressed the issue
                      I.   INTRODUCTION
                                                                       of developing a large scale search engine such as google but
    Road accidents are the major problem in many countries. It         failed to address the issue of specialized search[1]Sunny Lam,
is a very series problem in the highways of India. Internet is         “The Overview of Web Search Engines,” addressed the issue
grown very large. As it is very large and the information is           of how the search engines find information in the Web and
scattered all around the world, search engines are the only            how they rank the pages according to the given query. It helps
medium through which the information can be accessed. But              people perform Web searching easily and effectively. But it
the relevancy of the search result is the major problem in             not address the issue of not getting the required information
search engines. Though popular search engines like Google              even after search[2].
perform well through their quality of page ranking algorithms
still it is true that many questions remains unanswered up to          Robert Steele, “Techniques for Specialized Search Engines”
their expected relevancy. Special purpose search engines are           addresses the issue of the need for specialized search
those search engines which attempt to answer those questions           engine.[3]Z Xiang, K.           Wober,      DR. Fesenmaier,
which are not answered or cannot be answered by an all                 “Representation of the Online Tourism Domain in Search
purpose search engines.                                                Engines,” Addresses the issue of increasing the search results
                                                                       in tourism domain using techniques.But failed to provide the
   This project is an effort to create a comprehensive special         lack of important information related to tourism on the web[4].
purpose search engine which will support with accurate                 Z Xiang, Bing Pan, K. Wober, DR. Fesenmaier, “Developing
responses with maximum possible relevancy till the very last           SMART- Search : A Search Engine to Support the Long Tail in




                                                                 264                              http://sites.google.com/site/ijcsis/
                                                                                                  ISSN 1947-5500
                                                            (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                      Vol. 8, No. 2, 2010
Destination Marketing,” address the issue of effective                   after the model has been built[18].Pachaivannan Partheeban,
organizing over the internet to support travel information               Elangovan Arunbabu, Ranganathan Rani Hemamalini, “Road
search. It didn’t address the issue of how to increase the               Accident Cost Prediction Model Using Systems Dynamics
availability of travel information[5].                                   Approach,” addressed the issue of reducing the cost of accident
                                                                         using developing model using systems dynamic approach. But
Karl W Wober, “Domain Specific Search Engines,” addresses                not addresses that will it really lead to accurate cost
the techniques involved in domain specific search. But doesn’t           prediction[19].
address the issue of how to implement the domain specific
search engine[6].YAN Hongfei, LI Jingjing, ZHU Jiaji, PENG
Bo, “Tianwang Search Engine at TREC 2005: Terabyte                                I. NEED FOR A SPECIALIZED SEARCH ENGINE
Track,” address the issue of large amount of data transfer. It
does not address the issue of improving the search results[7].              All purpose search engines are very broad and deemed to
Jermy Ginsberg, Mathew H. Mohebbi, Rajan S. Patel,                       cover almost all domains in the world. Though this quality is
Lynnette Brammer, Mark S. Smolinski & Larry Brilliant                    an advantage it includes some inabilities too. The main factors
“Detecting influenza epidemics using search engine query                 which influence any search engine and create the specialized
data,” address the issue of detecting influenza using the query          need are found and listed as below:
data.It does not discuss about how avoid such epidemics using
that data[8].Gang Luo, Chunqiang, Hao Yang, Xing Wei,                          Specialization
“MedSearch : A Specialized Search Engine for Medical                           Availability of Information
Information,” addresses the issue of how to help layman in                     Responsibility
medical search but not addresses the issue of relevancy among                  Time elapsed
the medical information results[9].
                                                                         A. Specialization
Jianhan Zhu, Jun Hong, and John G. Hughes, “Using Markov                    Though all purpose search engines support specialization of
Chains for Link Prediction in Adaptive Web Sites,” addresses             information in response to the user queries, but they are
the navigation problems in adaptive web sites. But it does not           mainly meant for generalization of information. Curious
address link prediction from the past state to future state[10].         search engines use the user queries which are unanswered or
Junghoo Cho, Hector Gracia Molina, Lawrence Page,                        not properly answered with expected relevancy to enhance
“Efficient Crawling through URL Ordering,” addresses the                 their system to answer well in feature. But at that point of time
issue of in what order the crawler should visit URLs. But not            when user expects the right answer to his specialized queries
addressed the issue of taking care of the missed URLs which              he won’t be able to get.
are not came in order of the crawler[11].
                                                                         B. Availability of Information
Junghoo Cho, Hector Gracia Molina, “The evolution of the                    All purpose search engines gather information from all
Web and Implications of an Incremental Crawler,”addresses                around the web. It has tons of information to serve the users. It
the issue of incrementally updating the index. But not                   will answer the maximum of the user queries. But it won’t be
addressed the issue of updating the indexes randomly                     able to answer all the queries. Because it doesn’t possess the
[12]Junghoo Cho, Hector Gracia Molina, “Parallel Crawlers,”              information by its own. These search engines will struggle in
address the issue of managing the indexing of ever growing               answering queries which requests in depth details within a
web. It doesn’t give the complete guidelines to construct                particular domain.
parallel crawlers[13].Sanjay Kumar Singh, Ashish Misra,
“Road Accident Analysis : A Case Study of Patna City,”                   C. Responsibility
addresses the issue of Road accidents in Patna city. But not                 All purpose search engine tries through all the means to
addressed any road safety measures[14].G D Jacobs, Amy                   respond well for the user query and as well as update its
Aeron Thomas, “A Review of Global Road Accident                          information repository well to keep it fit for this activity. But it
Fatalities,” addresses the issue of deaths and injuries during           bears no responsibility to answer the queries positively. Hence,
accidents. But not addressed how public and private sector can           it is not sure for the user that his queries will be answered. It
act to prevent these injuries[15].                                       will be a trial and error process for him. All purpose search
                                                                         engines works with probability not with accuracy in this
    Ramasamy. N, “Accident Analysis of Chennai City,”                    aspect. Some search engine may handle some searches with
addressed the issue of accident analysis of Chennai city. But            most probably high relevancy and for some other searches
not addressed how to avoid such accidents in future[16].Dinesh           with less probability. This makes the user difficult to rely on
Mohan, “Social Cost of Road Traffic Crashes in India,”                   such kind of search engines.
addressed the issue of cost of injuries and deaths. But not
addresses how to eliminate those unwanted cost[17].P.Pramada             D. Time elapsed
VALLI, “Road Accident Models for Large Metropolitan Cities                  Time elapsed in searching is the major factor which affect
of India,” addressed the issue of preventing accidents by road           the interest of the user. When the time elapsed is more, it will
accident model. But not addresses how to avoid accidents even



                                                                   265                               http://sites.google.com/site/ijcsis/
                                                                                                     ISSN 1947-5500
                                                               (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                         Vol. 8, No. 2, 2010
create a greater amount of dissatisfaction in users. It has been                         VT5 – Two wheelers
found that the users are spending hours or sometime days in                              VT6 – Three wheelers
searching some essential information among the web. After                                VT7 – Others [bye cycle, bullock cart etc.,]
getting dissatisfied by their prolonged search they use to try
some other means to get that information i.e., making a series              Type of accidents:
of phone calls, trying in yellow pages, physically going to the
concerned place to get that information etc. Hence, the efforts                          F1 – Fatal (Death)
made to reduce this time elapsed will bring a giant leap in the                          F2 – Grievous Injury
development of the advanced search engines.                                              F3 – Minor Injury
                                                                                         F4 – Non Injury
The specialized search engine is aimed to address the above
factors which are not addressed by the all purpose search                   Time of accidents:
engines. Firstly it will concentrate on one domain and will have
sufficient collection of information to answer all sorts of                 Peak hours:
queries in that particular domain. As it is assured to answer all
the queries within that domain the user can fully rely on it.                             tp1 – 8:30 AM to 9:30 AM
Hence, it creates the full dependability to the user. The                                 tp2 – 5 PM to 6:30 PM
specialized search engine AccSearch will contain all needed
information local to its domain. It will ensure the availability of         Normal hours:
all the essential information. It bears the responsibility for the
information availability. It makes the user queries will be
                                                                                          tn1 – 10 AM to 5 PM
answered with full relevancy. It reduces the time elapsed in
searching by the user. It will answer the very first query itself                         tn2 – 7 PM to 8 AM (Cargo)
with full relevancy(whereas normally it needs many queries to               B. The Process Flow of Accident Analysis
obtain an information in an all purpose search engine). At
maximum level he may need to try with very few queries.
Finally he can finish off his search in few minutes instead of                                                  Traffic
long time hassles. It has been found that there are regular users                                              Authority
to search engine and they need to search for information for
their day to day activities. We identified the target users for
AccSearch. They are Police, NGOs, Statistical Bureaus,
Lawyers, Students to name a few. There will be bundle of
global users too. Once it attained perfection on its domain it                                               Highways in
will be made to crawl the whole www so that it will work                                                       Chennai
specialized on its domain and generalized on all-purpose search

                                                                                                             Vehicles on
II. MODIFIED MARKOV CHAIN ALGORITHM FOR ACCSEARCH                                                             the Road


A. Assumptions                                                                      Four or Six                                   Two or Three
                                                                                     wheelers                                       wheelers
       Types of vehicles : VT1, VT2, VT3, VT4, VT5, VT6, VT7                                            Select Type
                                                                                                          of Vehicle
       Types of accidents: F1, F2, F3,F4
       Time of accidents: tn1, tn2, tp1,tp2
                                                                                                                        Other
       Number of accidents : Ni
                                                                                                                       Vehicles
       Search engine : S1, S0
                                                                                    Number of                Number of               Number of
                                                                                    Accidents in             Accidents in            Accidents in
Types of vehicles:
                                                                                      a year                   a year                  a year
         VT1 – Government Bus
         VT2 – Private Bus                                                         Classification           Classification         Classification
         VT3 – Truck/Lorry                                                         of Accidents             of Accidents           of Accidents
         VT4 – Car/Jeep/Taxi/Tempo
                                                                                                        Accident Analysis


                                                                                    VT1       VT2      VT3         VT4       VT5     VT6        VT7
                                                                      266                                      http://sites.google.com/site/ijcsis/
                                                                                                               ISSN 1947-5500
                                                            (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                      Vol. 8, No. 2, 2010
   The equations which are the result of the accident analysis          Else If
are given below. Each element in an equation represents the
percentage of the type of accident occurred with respect to the         S1.Vehicle = VT6
total number of accidents by a particular vehicle type.                         &
                                                                         S1.Time = tp1 & tp2
     VT1  21.69F1 + 0.39F2 + 51.67F3 + 26.23F4
                                                                         Type.Accident = S0.[7.92F1 + 0.93F2 + 80.18F3 + 10.95F4]
     VT2  16.43F1 + 0.68F2 + 48.63F3 + 34.24F4
     VT3  19.74F1 + 0.31F2 + 42.63F3 + 37.3F4                          Else If
     VT4  7.27F1 + 1.07F2 + 58.37F3 + 33.27F4
                                                                         S1.Vehicle = VT7
     VT5  11.31F1 + 1.38F2 + 81.14F3 + 6.15F4                                 &
                                                                         S1.Time = tp1 & tp2
     VT6  7.92F1 + 0.93F2 + 80.18F3 + 10.95F4
     VT7  32.46F1 + 1.15F2 + 47.53F3 + 18.84F4                          Type.Accident = S0.[32.46F1 +1.15F2 + 47.53F3 + 18.84F4]
   As the vehicle types VT1 and VT2 are similar types (Bus)
                                                                        End If
and the VT2 is available in only negligible amount and its
effect on the accidents is very low these two types can be
merged.                                                                 III. MATHEMATICAL MODEL OF THE ALGORITHM
     VT1 & VT2  19.6F1 + 0.535F2 + 50.15F3 + 30.235F4
                                                                          The transition matrix has been constructed using these
                                                                        available results.
C. Algorithm

     Now the algorithm may be expressed as follows:                     V1  VT1 & VT2  19.6F1+0.535F2 + 50.15F3 + 30.235F4
                                                                        V2        VT3  19.74F1 + 0.31F2 + 42.63F3 + 37.3F4
If
     S1.Vehicle = VT1 & VT2                                             V3        VT4  7.27F1 + 1.07F2 + 58.37F3 + 33.27F4
           &
     S1.Time = tp1 & tp2                                                V4 VT5&VT6&VT717.23F1 +1.153F2+69.617F3+ 11.98F4

     Type.Accident = S0.[19.6F1+0.535F2+ 50.15F3 +30.235F4]             A. Transition Matrix

Else If                                                                                             V1          V2         V3         V4

     S1.Vehicle = VT3                                                                    F1       .196       .1974      .0727      .1727
           &
     S1.Time = tn2                                                                       F2        .00535 .0031          .0107      .01153
                                                                             T =
     Type.Accident = S0.[19.74F1 + 0.31F2 + 42.63F3 + 37.3F4]                                      .5015      .4263      .5837      .69617
                                                                                        F3
Else If                                                                                             .30235 .373           .3327      .1198
                                                                                        F4
     S1.Vehicle = VT4
           &
     S1.Time = tp1 & tp2

     Type.Accident = S0.[ 7.27F1 + 1.07F2 + 58.37F3 + 33.27F4]            The above traffic prediction analysis expressed in terms of
                                                                        transition matrix shows that the row represents fatality factors
Else If                                                                 in correspondence with the vehicle classification.

     S1.Vehicle = VT5                                                   In the same manner the columns of the matrix shows that
           &                                                            according to the vehicle types the percentage of fatality
     S1.Time = tp1 & tp2                                                occurred. As this is a real time classification which have been
                                                                        made in the Chennai city in the year 2008.
     Type.Accident = S0.[ 11.31F1 + 1.38F2 + 81.14F3 + 6.15F4]



                                                                  267                              http://sites.google.com/site/ijcsis/
                                                                                                   ISSN 1947-5500
                                                     (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                               Vol. 8, No. 2, 2010
  IV. EXPERIMENTAL VALIDATION AND RESULTS

                                                                                      Predictions for V3
                 Predictions for V1                                   0.7
  0.7
                                                                      0.6
  0.6
                                                                      0.5
  0.5
                                                                      0.4
  0.4
                                                                      0.3
  0.3
                                                                      0.2

  0.2
                                                                      0.1

  0.1
                                                                   6E-16
                                                                                 F1           F2             F3            F4
6E-16
                                                                     -0.1
            F1          F2          F3         F4
 -0.1                                                                       V3 (2010)         V3 (2011)           V3 (2012)

        V1 (2010)       V1 (2011)        V1 (2012)                          V3 (2013)         V3 (2008)

        V1 (2013)       V1 (2008)



                                                                                      Predictions for V4
                 Predictions for V2
                                                                      0.7
  0.7

                                                                      0.6
  0.6

  0.5                                                                 0.5

  0.4                                                                 0.4

  0.3                                                                 0.3

  0.2
                                                                      0.2
  0.1
                                                                      0.1
6E-16
            F1          F2          F3         F4                  6E-16
 -0.1                                                                            F1           F2             F3            F4

        V2 (2010)       V2 (2011)        V2 (2012)                   -0.1

        V2 (2013)       V2 (2008)                                           V4 (2010)         V4 (2011)           V4 (2012)
                                                                            V4 (2013)         V4 (2008)




                                                         268                               http://sites.google.com/site/ijcsis/
                                                                                           ISSN 1947-5500
                                                           (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                     Vol. 8, No. 2, 2010
                                                                        that 2011-2013 are nearby values and 2008 and 2010 are
  Predictions for V1                                                    distinct values. The values of 2011-2013 are 26.58%, 27.18%,
                                                                        and 27.09% respectively. The values of 2010 and 2008 are
   Predictions of F1 of 2010 – 2013are less when compared to            30.25% and 11.98%.
2008 whereas 2008 shows a value of 19.6% but 2010-2013
are found as 9.07%, 10.93%, 11.49% and 11.52%
respectively. In predictions of F2 2008 shows a lower value
that is 0.53% whereas 2010-2013 shows higher value i.e,                             V. CONCLUSION AND FUTURE WORKS
0.99%,1%, 1.02%, and 1.03% respectively. In the predictions
of F3 2010, 2012 and 2013 shows the higher values and the                  This paper presents AccSearch, a specialized web search
remaining shows the lower value. The values for 2010-2013               engine for road accident information retrieval. It will aid the
are 60.38%, 60.37 and 60.51 respectively. Similarly for 2011            user group consisting of police, NGOs, statistical bureaus,
and 2008 the values are 58.61% and 50.15%. Predictions of F4            lawyers, students and others who may require road accident
states that the value of 2008 is somewhat raised and 2010-              information for their day to day activities. AccSearch is
2013 are somewhat lowered. The values are 26.43%, 26.37%,               designed to be a scalable search engine. The primary goal is to
27.11% and 27.24% respectively whereas the 2008 value is                provide a very high relevancy in search results.
30.23%.
                                                                           In future this search engine will be enhanced as a semantic
  Predictions for V2                                                    search engine by creating ontology for this domain.

   Predictions of F1 of 2011-2013 are showing nearby values
where 11.53%, 11.45%, and 11.45% respectively whereas the
values of 2010 and 2008 are different those are, 13.47% and             ACKNOWLEDGMENT
19.74% respectively. Predictions of F2 shows that the values of           This paper kindly acknowledges the Traffic Police,
2010 -2013are almost similar, those are 0.99%, 1.01%, 1.02%,            Chennai,Tamil Nadu,India with whose support was very vital
and 1.02%. But 2008 shows 0.31%. Predictions of F3 shows                and acknowledges the institution where the idea was nurtured.
that 2010 , 2012 and 2013 shows almost similar values those
are 60.88%, 60.88% and 60.16%. The values of 2011and 2008
showing distinct such as 59.94% and 42.63%. Predictions of              REFERENCES
F4 shows that 2010-2013shows almost similar values those are
24.74%, 27.66%, 26.9%, and 27.1%.                                          [1]. Sergey Brin and Lawrence Page, “The Anatomy of
                                                                           Large-Scale Hypertextual Web Search Engine”, Computer
  Predictions for V3                                                       Networks 30(1-7): 107-117, 1998.
                                                                           [2]. Sunny Lam, “The Overview of Web Search Engines,”
   Predictions of F1states that 2010-2013 have almost similar                O. Waterloo – University of Waterloo, 2001.
values that is 11.63%, 11.44%, 11.5%, and 11.46%                           [3]. Robert Steele, “Techniques for Specialized Search
respectively whereas 2008 represents 7.27%. Predictions of F2                Engines,” Proceedings of Internet Computing, 2001.
is showing that 2011-2013 are same and 2010 is almost same                 [4]. Z Xiang, K. Wober, DR. Fesenmaier, “Representation
that is 1.02 for 2011-2013 , 1.05 for 2010 and 1.07 for 2013.                of the Online TourismDomain in Search Engines,” 47(2)
Predictions of F3 states that values of 2010-2013 are almost                 137 Journal of Travel Research, 2008
same those are, 61.33%, 60.18%, 60.25%, and 60.16%. For                    [5]. Z Xiang, Bing Pan, K. Wober, DR. Fesenmaier,
2008 it is 58.37%. Predictions of F4 states that the values of               “Developing SMART- Search : A Search Engine to
2010-2013 are almost same those are 26%, 27.43%, 27.07%                      Support the Long Tail in Destination Marketing,”
and 27.09% respectively. But the value of 2008 bears 33.27%.                 www.ota.cofc.edu.
                                                                           [6]. Karl W Wober, “Domain Specific Search Engines,”
                                                                             Wallingford,      UK        :     CABI       ,   2006,
Predictions of V4                                                            www.tourism.wu.wien.ac.at.
                                                                           [7]. YAN Hongfei, LI Jingjing, ZHU Jiaji, PENG Bo,
  Predictions of F1 states that the values of 2011-2013 are                  “Tianwang Search Engine at TREC 2005: Terabyte
almost similar values say 11.75%, 11.39%, and 11.49%. The                    Track,” Network and Distribution System Laboratory,
year 2010 which is 10.74%. 2008 shows a value 17.27%.                        School of Electronic Engineering and Computer Science,
Predictions of F2 states that the values of 2012 and 2013 are                Peking University Beijing, China.
similar and 2011 is almost similar, 1.02%, 1.02% and 1.03%.                [8]. Jermy Ginsberg, Mathew H. Mohebbi, Rajan S. Patel,
2010 shows 9.8% and 2008 shows 1.153%. Predictions of F3                     Lynnette Brammer, Mark S. Smolinski & Larry Brilliant
states that 2011-2013 and 2008 have almost similar values                    “Detecting influenza epidemics using search engine
these have 60.79%, 59.99%, 60.25%, and 69.617% whereas                       query data,” Nature, vol. 457, 19 February 2009.
2010 shows a lower value 58.13%. Predictions of F4 shows




                                                                  269                              http://sites.google.com/site/ijcsis/
                                                                                                   ISSN 1947-5500
                                                                    (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                              Vol. 8, No. 2, 2010
  [9]. Gang Luo, Chunqiang, Hao Yang, Xing Wei,
                                                                                        Fatal is shown as (2005: 584) (2006 : 627) (2007 : 583)
    “MedSearch : A Specialized Search Engine for Medical
                                                                                        (2008:612)(2009:609)
    Information,” IBM Research Journal, RC24205, March 6                                Non fatal is shown as (2005:4427) (2006 :4657) (2007: 4277)
    2007.                                                                               (2008:5774)(2009:4575)
  [10]. Jianhan Zhu, Jun Hong, and John G. Hughes, “Using
                                                                                        Where the total number of accidents of these categories are: 5177
    Markov Chains for Link Prediction in Adaptive Web
    Sites,” Soft-Ware 2002, LNCS 2311, pp 60 – 73,
    Springer-Verlog 2002.                                                        2.     Among the various vehicle types two wheelers is the more prone to
  [11]. Junghoo Cho, Hector Gracia Molina, Lawrence Page,                               accidents of fatal category. Lorrys are of the second categories
    “Efficient Crawling through URL Ordering,”                                          slightly less prone to accidents and so on.


  [12]. Proceedings of Seventh International Web                                                             2005          2006       2007      2008        2009
                                                                                        Two wheeler          154           174        159       191         191
    Conference (WWW 98), 1998.                                                          Lorry                108           120        113       104         103
  [13]. Junghoo Cho, Hector Gracia Molina, “The evolution                               MTC Bus              71            74         66        79          74
    of the Web and Implications of an Incremental Crawler,”                             Car                  60            72         75        75          77
    Department of Computer Science, Stanford University,                                Van                  63            57         59        52          38
    Stanford, CA 94305, USA.                                                            Auto                 48            48         40        41          36
                                                                                        UKV                  35            37         36        46          40
  [14]. Junghoo Cho, Hector Gracia Molina, “Parallel                                    Others               20            12         5         4           12
    Crawlers,” Department of Computer Science, Stanford                                 Private bus          14            22         15        22             19
    University, Stanford, CA 94305, USA.                                                Govt. bus            10            7          5         15              9
  [15]. Sanjay Kumar Singh, Ashish Misra, “Road Accident
    Analysis : A Case Study of Patna City,” Urbon Transport
                                                                                 3.     Similarly vehicle types are classified based on non fatal injuries
    Journal 2(2): 60-85, 2001.
                                                                                        on accidents
  [16]. G D Jacobs, Amy Aeron Thomas, “A Review of
    Global Road Accident Fatalities,” RoSPA Road Safety                                                     2005          2006      2007     2008      2009
    Congress, Plymouth, UK , 3-7 March 2000, at                                          Two wheeler        1501          1370      1373     1517      1438
    http://www.transport-links.org.                                                          Car            903           966       1065     1694      1229
  [17]. Ramasamy. N, “Accident Analysis of Chennai City,”                                   Auto            720           695       669      664       493
                                                                                            Lorry           473           425       375      654       456
    Working Paper 3, Centre for Road Safety, Central                                         Van            433           384       359      590       442
    Institute for Road Safety, Pune 2001.                                                 MTC Bus           184           202       274      334       285
  [18]. Dinesh Mohan, “Social Cost of Road Traffic Crashes                               Private Bus         63            59        61      129        68
    in India,” Proceedings : First Safe Community                                          Others            59            55        21       59        50
    Conference on Cost of Injury, pp 33-38, Viborg,                                         UKV              44            47        40       46        62
                                                                                            Jeep             29            23        22       26        29
    Denmark, October 2002.
                                                                                          Govt. bus          18            16        18       38        23
  [19]. P.Pramada VALLI, “Road Accident Models for Large
    Metropolitan Cities of India,” IATSS Research, Vol.29,
    No.1, 2005.
  [20]. Pachaivannan Partheeban, Elangovan Arunbabu,
    Ranganathan Rani Hemamalini, “Road Accident Cost
    Prediction Model Using Systems Dynamics Approach,”
    ISSN 1648-4142 print / ISSN 1648 – 3480 online 2008,                              4.Number of deaths as per the victim and as per the death:
    at www.transport.vgtu.it
                                                                                                                   2005      2006     2007    2008     2009
  [21]. “Road Safety Policy,” Road Safety Book, Home,                                      PEDESTRAIN               220       247      222     231      242
    Prohibition and Excise Department, Government of
    Tamil Nadu April 2007.                                                                  MCRIDER                148       200      211     236      202
                                                                                            CYCLIST                74        65       51      48       50
  [22]. Leslie Hogben, Matrices,Digraphs,Markov Chains &                                   MCPRIDER                45        35       41      36       43
                                                                                            OTHERS                 16        11       12       7       14
    Their Use by Google , Bay Area Mathematical
                                                                                          AUTODRIVER               11         6        9       7        4
    Adventures, Iowa State University and American                                       AUTOOCCUPANT               9        17       13       7        7
    Institute of Mathematics, February 27, 2008.


ANNEXURE:


  1.   Among the total number of accidents fatal accidents are low in
       numbers and the non-fatal are high in numbers. A sample trend is
       shown below:




                                                                          270                                 http://sites.google.com/site/ijcsis/
                                                                                                              ISSN 1947-5500
                                                                     (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                               Vol. 8, No. 2, 2010
            5.Number of injuries as per the victim
                 2005      2006      2007    2008          2009
   MCRIDER       1741      1785      1907    1927          1743
  PEDESTRAIN     1418      1335      1297    1373          1270                             10. Number of fatal injuries based on the road:
   MCRIDER       1741      1785      1907    1927          1743                       ROAD                 2005        2006        2007       2008   2009
    CYCLIST       493      407       347      347          241                       100FeetRoad            55           38         34         43     43
AUTOOCCUPANT      314      251       301      275          208                        OMR Road              38           35         33         52     44
  AUTODRIVER      204      178       180      183          150                        ECR Road              47           51         30         28     40
   STANDPER       172      147       185      170          129                        AnnaSalai             35           42         32         38     33
CAROCCUPANT        91       83       104      146          107                        ArcotRoad             23           17         16         26     16
  CARDRIVER       102       83        97      121          120                       200FeetRoad            14           24         15         16     15
 LORRYDRIVER       24       15        19       16           15                  ThiruvotriyurHighRoad       18            0          0          0    103
VANOCCUPANT        51       23        25       31           29                         SPRoad               11            6          9         11      8
  VANDRIVER        33       28        19       23           19                     Tharamani Road            9            7         13          8      6
                                                                                 VelacherryMainRoad           8         14          14          18   11
      6. Number of fatal injuries as per the age for fatal male:                    SNChettyStreet           10         12           4          11    8
    AGE        2005         2006       2007       2008     2009                  EnnoreExpressRoad            5          9           8          12   12
  15 to 29         130       149        182        169      11                  DurgabaiDeshmulkRoad          5          6           6           4    4
                                                                                   NewAvadiRoad               4          6          11           6    4
  30 to 44         131       139        117        141     116
                                                                                PoonamalleeHighRoad           0          9           3           2    4
  45 to 59         127       147        107        132     127
 ABOVE60           88         89         83        81       99
 BELOW14           16         15         12         7       11




        7. Number of non-fatal injuries as per the age of male:
    AGE            2005        2006         2007       2008        2009
   15 to 29        1387        1364         1440       1398        1217
   30 to 44        1254        1180         1170       1241        1044
   45 to 59         793         695          765        783         713
  ABOVE60           258         255          282        287         272
  BELOW14           219         196          213        193         184

         8. Number of fatal injuries as per the age of female:
     AGE          2005         2006         2007        2008       2009
    45to59          32          33           18          25         127
  ABOVE60           28          21           28          41         29
    30to44          20          18           23          13         17
  15to29            17          18           17          12         16
  BELOW14            7          10           10           6          6


       9. Number of non-fatal injuries as per the age of female:
     AGE           2005        2006        2007        2008        2009
  30to44            306         283         289         291         286
    15to29          287         253         242         278         210
    45to59          229         200         255         240         215
  ABOVE60           128         129         130         165         137
  BELOW14           101         101          86          96         102




                                                                          271                                 http://sites.google.com/site/ijcsis/
                                                                                                              ISSN 1947-5500