The paper 1 describes the associationrule mining, its classifications and the atmospheric components like roadwaysurface, climate, and light condition do not strongly influence the fatal accidentrate. But the human factors like being alcoholic or not, and the impact havestrongly affect on the fatal accident rate.
Acommon mechanism to recognize the relations between the data stored in hugedatabase and plays a very significant role in repeated object set mining isassociation rule mining algorithm. A classical association rule mining methodis the Apriori algorithm whose main aim is to identify repeated object sets toanalyze the roadway traffic data. Classification in data mining methodology focusat building a classifier model from a training data set that is used toclassify records of unrevealed class labels. The Naïve Bayes technique is oneof the probability-based methods for classification and is based on the Bayes’hypothesis with the probability of self-rule between every set of variables. The author applies statisticsanalysis and Fatal Accident Reporting System (FARS) to solve this problem. Fromthe clustering result some regions have larger fatal rate but some others havesmaller. When driving within those risky or dangerous states, people take moreattention.
When the task performed, data seems never to be sufficient to make astrong choice. If non-fatal accident data, weather condition data, mileagedata, and so on are available, more test could be executed thus more advicecould be made from the data.In paper 2, we proposed aframework that is used K-modes clustering technique as an initial work fordivision of various types of road accidents on road network. Then association rulemining are used to recognize the different situations that are related with theoccurrence of an accident for both the entire data set (EDS) and the clustersrecognized by K-modes clustering algorithm. Six clusters (C1toC6) are used basedon properties accident type, road type, lightning on road and road featureidentified by K modes clustering method. On each cluster association rulemining is applied as well as on EDS to create rules.
Powerful methods with higherraise values are taken for the inspection. Rules for various clusters disclose thesituations related with the accidents within that cluster. These rules are comparedwith the rules created for the EDS and resemblance shows that association rulesfor EDS does not disclose correct information that can be related with anaccident. More information can be identified if more feature are presented thatis associated with an accident. To buildup our methodology, we also performedtendency analysis of all clusters and EDS on monthly and hourly basis. Theresults of analysis assist methodology that performing clusteringprior to analysis helps in identify better and useful results that we cannotobtained without using cluster analysis