A Study On Crime Prediction And Detection Techniques Using Data Mining

Data Mining is the technique from where we can learn to create design and pattern from a huge data set. It can help us in various fields like future health care, cancer detection and survival, educational development, child behavior and development, crime detection and fraud analysis etc. Among all crime prediction, crime detection, fraud analysis are some interesting topics because in the recent past there had been huge increase in the crime. It directly affects our national security significantly. By using the applications of data mining we can learn the crime characteristics to help the society and develop our morality. This paper is a study on crime prediction and detection techniques using data mining which can help to design new strategies to analysis crime.

Introduction

Mining any data is the techniques of dealing with expensive unordered, unsorted and raw data indexes to summarize the data and get information that can help in future. Data mining is a procedure to analyze data from an informational collection. It predicts future patterns and additionally allows the organization to form the training driven call. Generally used techniques for mining of data or knowledge are artificial neural networks, rule induction, decision tree, NN technique, genetic algorithm, Machine learning. They're applied in several fields. One such fascinating application is crime investigation. A criminal offense is associate with unlawful activity that a man will be punished by law. Crime against someone is termed as personal crime like murder, robbery, etc. Property crime means that felony of property. Web crime is also a big problem nowadays. Web crime is the use of internet based attacks in crime activities, including acts of deliberate, large scale disruption of computer network, especially of the personal computer attach to the internet by the means of tools such as viruses. Crime analysis could be a law implementation task which has associate degree organized analysis that acknowledges and determines the pattern of crime. Crime will be classified into differing types however, in this, we tend to targeted on types of crimes i.e. Fraud detection, traffic violence, violent crime, web crime and sexual offense. The assorted techniques used for various crimes are mentioned with associate degree introduction to the involved crime.

Types of crime

Criminal activity has invariably been a priority for governments round the world, several researches has been done to effectively realize countermeasures and indicators of crime before happening. Criminals are major attention from enforcement organizations.

Crimes that are happening around us can be of many types. Crime can be murder or attempt to murder, robbery of a bank, web crime like fraud detection or pornography, accident or violence of traffic rules. We will discuss about four main types of crime and use of data mining in predicting and preventing the crimes.

Web Crime

Web crime or cyber crime has the same meaning. It is mainly internet based attack that harms people by stealing someone’s personal information or harassing someone over internet or using pornography over the network. Nowadays this crime has taken place to large volume and victims are suffering for the wrong usage of internet. Web crime analysis should be incredibly vital responsibility of law enforcement organization in any country.

Traffic Violence

When the driver of a vehicle breaks the law of traffic that damage or may damage others is called traffic violation. In our country recently there are many accidents that caused death have increased in numbers that is very alarming. These caused only because the drivers and the general people are very clumsy about the rules and regulation for the traffic also sometimes the police are also not attentive at their duty. To solve the above mentioned problems so prevention must take.

Fraud Detection

Taking an advantage from another which is not fair or misdirecting people by giving false information is called fraud. A fraud incorporates such activities that are illegal and can harm others. There are various kind of fraud activities and they include credit card fraud, web sale fraud, insurance card fraud, check fraud, credit card and master card fraud etc. Check fraud means that provision of a check once enough cash is not transfer in the specific account; web sale means that saling false items; insurance fraud means that wrong or false insurance claimed for automobile harm, any credit card and MasterCard fraud means that getting data from varied means that that is employed for big quantity of purchase while not have authentication.

Violent Crime

Violent crimes are that kind of crime where guilty party threaten or flew after the crime. These crimes involve killing someone that is murder or attempt to murder. Another one is rape. These are two act of violent crime which is still happening trough out the world.

III. Detections of various crimes

Techniques for Fraud Detection

There are several techniques which are used to detect any fraud. Some techniques are discussed here below: Genetic Algorithm is one of them. It is mainly structured on biological choice process. Normally at each iteration it creates points. It associates with degree of optimum resolution. It is associated with the tremendous category of transformative algorithms that creates solutions to improve problems like inheritance. Application of this algorithmic program square measure bioinformatics, process science, engineering, arithmetic, physics and different fields. Advantage of generic algorithmic program square measure more economical, faster, optimize continuous and separate operate. The steps of this algorithm are described below:

1. Selection: Chromosomes are selected from a group.
2. Crossover: With a hybrid likelihood, main set of data form a new set of data. If no hybrid event was performed, new set of data is the copy of the main set of data.
3. Mutation: Likelihood of new production at each locus.
4. Accepting: New data are placed in the new sets. HMM is another technique which is called Hidden Markov Model. It can be an applied as mathematics model of method consisting two random variables that modify the state consecutive. It's restricted arrangement of states in administered set of modification possibilities. It allows North American country to discuss regarding each monitored events and hidden event.

General Techniques in Traffic violence

Cumulative Logistic Model predicts likelihood of the chosen response in normal scale. The assumption of its proportional odds i.e. Coefficients of predictor class which is consistent all told levels of response. Association rule mining may be a procedure of finding random patterns, correlations, associations, or causative structures of data sets. Data sets are obtainable in numerous varieties of information like relative databases, transactional databases etc. K Means Clustering is unsupervised learning. Here we group data based on their similarity. Similarity is measured by their distance. Initially K centroids are chosen randomly for K clusters. Then data points are set to the nearest cluster. Advantage of k mean rule area unit quick, robust, simple to know and provides best result. For portioning the cluster there are a unit alternative ways Steps of K means clustering are

• The algorithm arbitrarily selects K points as the initial cluster centers.
• Each point in the data set is assigned to the closed cluster based on the Euclidean distance between each point and each cluster center.
• Each cluster center is recomputed as the average of the points in that cluster.
• Steps 2 and 3 repeats until the clusters converge.

Neural Network is an efficient way in pattern recognition. It involves inter-connected network of comparatively straightforward and usually non-linear unit. This formula is non-algorithmic and recording machine strategy that is trainable. This algorithm is engaging to PR system designer because it needs previous and detail information of internal system operation is lowest. Application of Neural network square measure character recognition and image processing. There are some characteristics the are described below:

• Network topology or interconnection of neural unit.
• Characteristics of individual unit or neuron.
• Strategy of pattern learning or training.
• High dimensional and complex interaction between problem variable.
• Feature extraction from complex dataset.

General Techniques in Web Crime Detection

K-Means Clustering Algorithm is mainly used in web crime data. Traditional data processing techniques like association analysis, classification, cluster analysis area unit used to determine patterns in structured knowledge. Their area unit newer techniques which will be wont to determine patterns in structured similarly as in unstructured knowledge. Entity extraction identifies explicit patterns of information like text, pictures and audio knowledge etc. entity extraction provides basic info for crime analysis, however its performance depends greatly upon the supply of excessive quantity of fresh computer file. Bunch techniques cluster knowledge things into categories with similar characteristics to maximize or minimize similarity. bunch crime incidents will automatize major a part of crime analysis however it's restricted by the high process intensity usually needed.

Related work

We know about the related work behind web crime data. Syed Ahsan Shabbier, described Generic algorithmic program for preventing credit card frauds. It had been used for up the computing value with time by making advanced systems. It might analyze a dishonest group action in few second. The chance of false statement exchanges might anticipate shortly when credit card exchanges and arrangement of hostile to extortion systems can be received to stay banks from unbelievable misfortunes and minimize dangers.

Naeimeh Laleh mentioned supervised strategies, semi-supervised strategies, unsupervised strategies, and real time approaches to sight the kind of fraud and compare the different techniques.

Chao Yang mentioned regarding rough-fuzzy c-means formula for analysis of violent crime, rough set and knowledge entropy. It had been combined to upgrade the capability so it could affect the uncertainty, vagueness, and wholeness. This formula was used for resolving overlapping information. Chao Yang planned swarm rough rule to analyze the combo parts of brutal crime and break down 3 varieties of combine factors, i.e. Genetic, natural and psychological factors and assessed the execution and therefore the fuzzy swarm improvement technique by obtaining numerous diminishments for the combo issue datasets. It works higher in an exceedingly mixed dataset cluster.

Anshu Sharma, planned k means that cluster algorithmic program that was used for constructing patterns of information. information were collected and distributed, 2 third of true information and misrepresentation history data were utilized for making ready and remaining data were utilized for forecast and net crime discovery. The exactness of the planned work was 94.75 try to it fruitfully recognized the false rate of five

K. K. Sindhu explained scientific investigation ventures within the capability media and hidden knowledge investigation within the record framework, network rhetorical and cyber-crime mining. Device was projected by combining digital rhetorical investigation and mining of crime knowledge intended for locating motive and pattern of attacks and hecks of assaults types occurred in that period.

Conclusion

Any type of crime is a crime whether it is small or big. Crime detection is important in today’s world of technology. With the growth of internet and technology the possibility of wrong doing is also increasing. Where some people is getting the good side of the internet there are some other people who misuse and mislead information. This paper shows how various types of crimes are detected and predicted. Here I looked at the use of data mining for identifying crime patterns using various techniques of data mining Also data mining is sensitive to quality of input data that may be inaccurate, have missing data, be data entry error prone etc. Additionally mapping real data to data mining attributes isn't perpetually a simple or easy doing task and sometimes requires competent knowledge laborer and crime data analyst with good domain information. They have to figure closely with a detective within the initial phases. In today’s world data mining has huge scope and various field. By using various predictive and descriptive model of data mining we can minimize the crime and occurrences that are happening around us.