discrimination prevention

Top PDF discrimination prevention:

DIRECT AND INDIRECT DISCRIMINATION PREVENTION AND RULE GENERATION IN DATA MINING

DIRECT AND INDIRECT DISCRIMINATION PREVENTION AND RULE GENERATION IN DATA MINING

Abstract: In social and economic science, discrimination is the subject which has been extensively studied in data mining. Discrimination can be categorized into direct and indirect. Decisions based on sensitive attributes are termed as direct discrimination and the decisions which are based on non-sensitive attributes are termed as indirect discrimination which is strongly correlated with biased sensitive once. There are many new techniques propose for solving discrimination prevention problems by applying direct or indirect discrimination prevention individually or both at the same time. New metrics to evaluate the utility were proposed and are comparing with approaches. The propose work discusses how privacy preservation and prevention between discrimination is implement with the help of post processing approach. The Classification Based on Multiple-class Association Rules (CMAR) is a kind of association classification methods which combines the advantages of both associative classification and traditional rule-based classification which is used to prevent discrimination prevention in post processing.
Show more

12 Read more

Discrimination Prevention by Different Measures in Direct Rule Protection Algorithm

Discrimination Prevention by Different Measures in Direct Rule Protection Algorithm

A. Discrimination Prevention by Pre-processing Method In this method, the original dataset is modified so that it will not result in discriminatory classification rule. In a preprocessing method any data mining algorithm can be applied to get mining model. It consists of data massaging step. Kamiran and Calder [2] proposed a method based on data massaging where class label of some of the records in the dataset is changed but as this method is intrusive, concept of Preferential sampling was introduced where distribution of objects in a given dataset is changed to make it non-discriminatory [2]. It is based on the idea that, "Data objects that are close to the decision boundary are more vulnerable to be victim of discrimination." This method uses Ranking function and there is no need to change the class labels. This method first divides data into 4 groups that are DP, DN, PP, PN, where first letters D and P indicate Deprived and Privileged class respectively and second letters P, N indicates positive and negative class label. The ranker function then sorts data in ascending order with respect to positive class label. Later it changes sample size in respective group to make that data biased free. Sara Hajian and Josep Domingo-Ferrer [4] proposed another preprocessing method to remove direct and indirect discrimination from original dataset. It employees ‘elift’ as discrimination measure to prevent discrimination in crime and intrusion detection system [5]. Preprocessing method is useful in applications where data mining is to be performed by third party and data needs to publish for public usage [4].
Show more

6 Read more

Direct and Indirect Discrimination Prevention Approach using Association Rule Hiding

Direct and Indirect Discrimination Prevention Approach using Association Rule Hiding

Sara Hajian and J. Domingo-Ferrer [8] have introduced anti-discrimination in the context of cyber security. They have developed a new discrimination prevention method based on data transformation that can consider several discriminatory attributes and their combinations. They have introduced some measures for evaluating this method in terms of its success in discrimination prevention and its impact on data quality. The limitation of this approach is that it deals only with direct discrimination.

9 Read more

Sensitive Attribute Discrimination Prevention in Data Mining

Sensitive Attribute Discrimination Prevention in Data Mining

In the year 2008, P Pedreschi et al, Discrimination prevention technique in [15] consists of inducing a classifier like Naive Bayes in which classification is done without any sensitive attribute. In this technique prevention need to modify probability of decision records. That does not lead to discriminatory decisions even if trained from a dataset containing these item set. Training model consist unwanted dependencies between attributes.In the year 2009, F Kamiran, et al, [20] had tackled the problem of impartial classification by introducing a new classification scheme for learning unbiased models on biased training data in 2009. The method is based on massaging the dataset by making the least intrusive modifications which lead to an unbiased dataset. Numerical attributes and group of attributes are not considered as sensitive attribute.
Show more

6 Read more

Direct and Indirect Discrimination Prevention in Data Mining

Direct and Indirect Discrimination Prevention in Data Mining

Our Proposed data transformation methods rule protection and rule generalization are based on measures for both direct and indirect discrimination and can deal with several discriminatory items. We demonstrate an integrated approaching to address and indirect discrimination prevention, on finalized algorithmic rule and all potential information shift ways confirmed rule protection and or convention generalization that could indirect discrimination prevention. We suggest fresh utility amounts to evaluate the different aimed favouritism prevention processes in terms by information quality and discrimination removal as some direct and indirect discrimination. Direct and indirect discrimination discovery includes identifying discriminatory rules and redlining rules.
Show more

6 Read more

Discrimination Prevention Mechanism with Differential Privacy Scheme in Data Mining

Discrimination Prevention Mechanism with Differential Privacy Scheme in Data Mining

Discrimination prevention process is designed to protect decisions. Rule generalization and rule prevention algorithms are enhanced for dynamic policy model. Direct and indirect discrimination prevention algorithm is also tuned for dynamic policy scheme. Discriminations are protected with reference to sensitive and non-sensitive attributes. One of these measures is the extended lift (elift). The purpose of direct discrimination discovery is to identify α-discriminatory rules. In fact, α– discriminatory rules indicate biased rules that are directly inferred from discriminatory items. We call these rules direct α-discriminatory rules. In addition to elift, two other measures slift and olift were proposed by Pedreschi et al. The reason is that different measures of discriminating power of the mined decision rules can be defined, according to the various antidiscrimination regulations in different countries. Yet the protection methods are similar no matter the measure adopted. To determine the redlining rules stated the theorem
Show more

5 Read more

A Methodology For Direct And Indirect Discrimination Prevention In Data Mining

A Methodology For Direct And Indirect Discrimination Prevention In Data Mining

Data mining is an increasingly important technology for extracting useful knowledge hidden in large collections of data. There are, however, negative social perceptions about data mining, among which potential privacy invasion and potential discrimination. The latter consists of unfairly treating people on the basis of their belonging to a specific group. Automated data collection and data mining techniques such as classification rule mining have paved the way to making automated decisions, like loan granting/denial, insurance premium computation, etc. If the training data sets are biased in what regards discriminatory (sensitive) attributes like gender, race, religion, etc., discriminatory decisions may ensue. For this reason, antidiscrimination techniques including discrimination discovery and prevention have been introduced in data mining. Discrimination can be either direct or indirect. Direct discrimination occurs when decisions are made based on sensitive attributes. Indirect discrimination occurs when decisions are made based on non sensitive attributes which are strongly correlated with biased sensitive ones. In this paper, we tackle discrimination prevention in data mining and propose new techniques applicable for direct or indirect discrimination prevention individually or both at the same time. We discuss how to clean training data sets and outsourced data sets in such a way that direct
Show more

5 Read more

A Methodology for Sensitive Attribute Discrimination Prevention in Data Mining

A Methodology for Sensitive Attribute Discrimination Prevention in Data Mining

ABSTRACT: Today, Data mining is an increasingly important technology. It is a process of extracting useful knowledge from large collections of data. There are some negative view about data mining, among which potential privacy and potential discrimination. Discrimination means is the unequal or unfairly treating people on the basis of their specific belonging group. If the data sets are divided on the basis of sensitive attributes like gender, race, religion, etc., discriminatory decisions may ensue. For this reason, antidiscrimination laws for discrimination prevention have been introduced for data mining. Discrimination can be either direct or indirect. Direct discrimination occurs when decisions are made based on some sensitive attributes. It consists of rules or procedures that explicitly mention minority or disadvantaged groups based on sensitive discriminatory attributes related to group membership. Indirect discrimination occurs when decisions are made based on non sensitive attributes which are strongly related with biased sensitive ones. It consists of rules or procedures that, which is not explicitly mentioning discriminatory attributes, intentionally or unintentionally, could generate decisions about discriminatio n.
Show more

6 Read more

DISCRIMINATION PREVENTION IN DATA MINING

DISCRIMINATION PREVENTION IN DATA MINING

Classification rules are actually learned by the system based on training data. If the training data is biased against a community, the learned model will also show a discriminatory behavior. If the original biased dataset is used for data analysis without any anti-discrimination process which is discrimination discovery and as well as discrimination prevention, the extracted rules which are discriminatory can lead to automated unfair decisions Determining such possible biases and removing them from the training data without harming their decision making utility is highly desirable. Data mining can be both a source of discrimination and a means for discovering discrimination. Direct discrimination consists of rules that mention minority or disadvantaged groups based on sensitive attributes that are discriminatory.
Show more

6 Read more

Feature Correlation Measure Based Real Time Discrimination Prevention with Transactional Data Sets using Social Networks

Feature Correlation Measure Based Real Time Discrimination Prevention with Transactional Data Sets using Social Networks

presents a multi dimension model which provides higher flexibility. The method uses a greedy algorithm to approximate with simple way. In [11], the author presents a discrimination prevention approach towards mitigating discrimination attack performed by a group of people. The method mitigates group threat and reduces the distortion In [12], a discrimination prevention technique is presented which handle the direct and indirect discrimination. The method uses outsourced data in generating nondiscriminatory rule to mitigate in direct and indirect rules. Similarly, [13] discuss the approach for privacy preservation. The method uses taxonomy in the preservation of user data. In [14], the author presents a discrimination prevention technique towards crime detection. The author discuss different cleaning algorithm to perform cleaning and uses outsourced datasets in such a way that legitimate classification rules can still be extracted but discriminating rules based on sensitive attributes cannot. All the above discussed methods has produces poor results in sanitization and discrimination.
Show more

5 Read more

Survey of Approaches for Discrimination          Prevention in Data Mining

Survey of Approaches for Discrimination Prevention in Data Mining

Kamiran F. et al.[12] have developed a new solution to the CND problem by introducing a sampling scheme for making the data discrimination free instead of relabeling the dataset. This approach is based on pre-processing method of discrimination prevention. It consists of sampling the data objects with replacement to make the dataset bias free. A Preferential Sampling (PS) scheme is introduced to make the dataset bias free. This approach deals with changing the distribution of different data objects for a given data to make it discrimination free. The idea is that the data objects close to the decision boundaries are more prone to be the victim of discrimination. The distribution of these borderline objects is changed to make the dataset discrimination free. To find the least certain elements, a ranking function, learned on the original data, is used to identify the data objects close to the borderline. Then, based on the sanitized data, a non-discriminatory model can be learned. Since this model is learned on non- discriminatory data, it reduces the prejudicial behaviour for future classification. It helps in obtaining good results with both stable and unstable classifiers. It mitigates the discrimination level by maintaining high accuracy level.
Show more

5 Read more

Index Terms Antidiscrimination, data mining, direct and indirect discrimination prevention, Rule protection,

Index Terms Antidiscrimination, data mining, direct and indirect discrimination prevention, Rule protection,

The determination of this paper was to progress a new pre-processing discrimination prevention methodology including different data transformation methods that is used to prevent direct discrimination, indirect discrimination or both of them. To attain this objective, firstly, they measure discrimination and identify categories and groups of individuals that have been directly or indirectly discriminated in the decision-making processes; secondly, have to transform data in a proper way to remove all those discriminatory biases. Finally, discrimination-free data models can be produced from the transformed data set without seriously damaging data quality. Proposed algorithm used to measure the quality and help to prevent direct and indirect discrimination.
Show more

5 Read more

Discrimination Prevention with Privacy Preservation in Data Mining

Discrimination Prevention with Privacy Preservation in Data Mining

the data so that discrimination is removed with minimum possible changes. The class labels of the most probable victims (discriminated ones) and profiteers (favored ones) was changed by them. Naive Bayesian classifier ranker is used to determine the victims and profiteers. This ranker ranks the data objects in accordance with their probability of being in the target class. The training data is modified until the data becomes discrimination free. It has disadvantages like it works only for direct discrimination and applicable to only one discriminatory attribute. The paper [3] proposed a method for indirect discrimination prevention which considers several discriminatory attributes. They first discovered whether there exists indirect discrimination. In the event of any indirect discrimination, the modification of dataset is done until discrimination is decreased to a certain threshold or is totally eliminated. The method consists of transforming the original data in such that each redlining rule gets converted to non-redlining rule with minimum loss of information. The transformed dataset is evaluated with measures like Discrimination Prevention Degree, Discrimination Protection Preservation. It has limits like, it works for only indirect discrimination and there is little information loss.
Show more

5 Read more

Implementation Of Privacy Preservation Using Anonymization Methods For Discrimination Prevention

Implementation Of Privacy Preservation Using Anonymization Methods For Discrimination Prevention

BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies) is an unsupervised data mining algorithm. BIRCH can be used to perform clustering in discrimination environment. It can be used in multi-dimensional datasets and it has minimized I/O cost than Apriori algorithm (1 or 2 scans). First, it scans the data set and construct clustering feature tree in its memory. Then it condenses large clustering feature tree into smaller one and performs global clustering by using its centroid points [20]. Finally it does cluster refining
Show more

5 Read more

IM LR: An approach for Direct and Indirect Discrimination Prevention

IM LR: An approach for Direct and Indirect Discrimination Prevention

The given research work essence on discernment anticipation built on pre-processing, since the pre-processing method look like the utmost the one: it does not need varying the customary data mining algorithms, unlike the in treating method, and it agree data issuing (rather than just knowledge publishing), unlike the post dispensation method. The suggest work overawed the restraint based on pre-processing publish so far. In the suggest work new data alteration approaches are based on procedures for both direct and indirect discernment and can deal with numerous discriminatory items. This suggests method assurance that the converted data set is surely discernment allowed. It consists of measure to assess how much discernment has been removed and how much info loss has been sustained. Hence, the suggest work method to discernment anticipation is broader than in preceding work. Recommend work current a unified method to direct and indirect discernment prevention, with confirmed algorithms and all probable data conversion procedures that could be applied for direct or indirect discernment anticipation also identify the dissimilar structures of every process. The proposition techniques established new metrics that identify which data should be reformed, how many data should be different, and how those records should be different during data transformation. In adding, novel utility methods to calculate the dissimilar suggested discernment preclusion procedures in terms of data superiority and discernment elimination for together direct and indirect discernment. Based on the planned measures, present widespread investigational effects and compare the different possible methods for direct or indirect discernment anticipation to find out which procedures could be more successful in relations of low info loss and high discrimination removal.
Show more

9 Read more

A REVIEW ON AUTOMATIC DISCRIMINATION IDENTIFICATION SYSTEM FOR CORPORATE COMPANIES

A REVIEW ON AUTOMATIC DISCRIMINATION IDENTIFICATION SYSTEM FOR CORPORATE COMPANIES

Abstract: Data mining is an increasingly important technology for extracting useful knowledge hidden in large collections of data. Automated data collection and data mining techniques such as classification rule mining have paved the way to making automated decisions, like loan granting/denial, insurance premium computation, etc. If the training data sets are biased in what regards discriminatory (sensitive) attributes like gender, race, religion, etc., discriminatory decisions may ensue. For this reason, antidiscrimination techniques including discrimination discovery and prevention have been introduced in data mining. Discrimination can be either direct or indirect. Direct discrimination occurs when decisions are made based on sensitive attributes. Indirect discrimination occurs when decisions are made based on no sensitive attributes which are strongly correlated with biased sensitive ones. We tackle discrimination prevention in data mining and propose new techniques applicable for direct or indirect discrimination prevention individually or both at the same time. Here we are developing an application in which employee’s loan, training and promotions depending on performance will be managed and also we are developing an application in which discrimination will be identified automatically and complaint will be posted against discriminator to the administrator.
Show more

6 Read more

Review on Discrimination Detection and Prevention in Data Mining

Review on Discrimination Detection and Prevention in Data Mining

On the request of the user the detection and prevention system fetches the data from original dataset. Before starting the detection and prevention the data needs to be pre-processed so as to feed as input to the further steps. The next step is that Frequent Classification Rules are applied on the pre-processed data for the purpose of classification among classes and attributes of the classes. The next step we can say is the actual step of working model i.e. Discrimination Discovery here the discrimination present in the hidden dataset is determined for the purpose of applying discrimination prevention technique, it is necessary to detect the type of discrimination present in the dataset. There are several methods of discrimination prevention as we discussed in literature review section of this paper here we are interested in preserving the quality of data along with making it bias free hence here we opted the method of data transformation we did not selected data sanitization or any other method for the same due to the loss of quality information. Once the data is transformed we get the resultant database as bias free, one can go for fair decision making process relying on the dataset we just obtained after data transformation. The next step is to apply any standard or special purpose mining algorithm for extraction of information. The final step is to return the discrimination free dataset to the user for the purpose of automated decision making or for publishing for public usage.
Show more

6 Read more

A RESULT ON AUTOMATIC DISCRIMINATION IDENTIFICATION SYSTEM FOR CORPORATE COMPANIES

A RESULT ON AUTOMATIC DISCRIMINATION IDENTIFICATION SYSTEM FOR CORPORATE COMPANIES

Discrimination prevention methods based on preprocessing published so far [1], [2] present some limitations, which we next highlight: They attempt to detect discrimination in the original data only for one discriminatory item and based on a single measure. This approach cannot guarantee that the transformed data set is really discrimination free, because it is known that discriminatory behaviors can often be hidden behind several discriminatory items, and even behind combinations of them.

7 Read more

REVIEW ON DISCRIMINATION DISCOVERY AND PREVENTION TECHNIQUES IN DATA MINING

REVIEW ON DISCRIMINATION DISCOVERY AND PREVENTION TECHNIQUES IN DATA MINING

Abstract: In data mining, discrimination is the subject which has been extensively studied in social and economic science. It is more than observable that the majority people do not want to be discriminated because of their gender, nationality, religion, age and so on. This problem mainly arises when these kind of attributes are used for decision making purpose such as giving them a job, loan. Insurance etc.. For this reason discovering such attributes and eliminating them from the training data without affecting their decision-making utility is essential. So we introduce an antidiscrimination techniques which including discrimination discovery and prevention. Discrimination prevention is mainly used for the purpose of inducing patterns that do not lead to discriminatory decisions even if the original training datasets contain any of the discriminatory attributes. In the discrimination prevention method, we introduce a group of pre-processing discrimination prevention methods and specify the different features of each approach and how these approaches deal with direct or indirect discrimination. We discuss how to clean training data sets and outsourced data sets in such a way that direct and/or indirect discriminatory decision rules are converted to nondiscriminatory classification rules.
Show more

8 Read more

Identification Of Discrimination And Prevention Methods In Data Mining

Identification Of Discrimination And Prevention Methods In Data Mining

The purpose of this paper was to develop a new preprocessing discrimination prevention methodology including different data transformation methods that can prove direct discrimination, indirect discrimination or both of them at the same time. To attain this objective, the first step is to measure discrimination and identify categories and groups of individuals that have been directly and/or indirectly discriminated in the decision-making processes .the second step is to transform data in the proper way to remove all those discriminatory biases. Finally, discrimination-free data models can be produced from the transformed data set without seriously damaging data quality.
Show more

5 Read more

Show all 10000 documents...