Chosen topic to be used in data science methodology
I have chosen credit cards. In the journey of data science, it is significant for one to think about ways in which data science can be useful when it comes to generating values that can be used to save time. Hence, through using the data set in Kaggle about fraud detection with a credit card, one gets hooked up with the use of credit cards in data science methodology.
Business understanding stage
Problem description
By the use of credit cards, it is possible to know the feelings which one has whenever his/her ticket is rejected as a result of suspected fraudulent transactions. In the year 2019, an industry report established that one out of six valid credit cardholders faced at least one case of declined transaction (Schroeder et al. 13). These arose due to imprecise fraud detection within the past years. That makes the detection of fraud to be a predicament deemed expensive for the users of these cards. The declined transactions contribute to approximately $119 in terms of loss every year (Schroeder et al. 12). Though multiple approaches of machine learning have been developed in curbing the fraud witnessed with the credit cards, there exist automation platforms that have been newly introduced to assist in the problem. Nowadays, firms are struggling to make investments in the automation process to stem down the fraud cases experienced in using credit cards. Without appropriate measures taken by organizations, fraudulent activities using a credit card would continue ravaging various operations of big firms leading to annual losses to soar. Thus, frauds experienced in using credit cards should be reduced through an available process to curb losses experienced. I would like to solve the problem of credit card fraud which is causing organizations to lose money annually.
Problem as a question to be answered using data
Through the use of data available on fraudulent activities committed using credit cards, one is left with multiple questions to answer. For example, can we establish the number of frauds committed using credit cards in various companies in a year?
Stages for the problem
Analytic approach- upon giving an outline of the business problem, the data scientist would be in a position to define analytic approach which can be used to solve the problem which has been described (Angelov, Plamen, Xiaowei Gu and Príncipe 2981). In the analytic approach, it will be possible to express the problem defined in statistical context and machine learning technique. Data scientist will be in a position to recognize techniques apt for attained desired results at the end.
Data requirements- appropriate selection of analytic approach establishes data requirements. In order to use an analytic method appropriately, then it will need certain data contents, formats as well as representations. This will be guided by the knowledge domain.
Data collection- with this problem, the data scientist will identify and gather data resources. The collected data in the form semi-structured, structured and unstructured collected must be pertinent to the problem domain. A continuous revision of data requirement will be done upon meeting gaps in the data collections. More data collected to assist in acquiring information to be used in solving the problem.
Data understanding and preparation- to understand the data, data scientist will be using descriptive statistics as well as visualization techniques in order to comprehend the content. Assessment of data quality in addition to discovering initial data insights shall be made. In order to close up the gaps, previous steps and data collection methods can be used to further comprehend the needed data(Angelov, Plamen, Xiaowei Gu and Príncipe 2983). Appropriate data preparation is paramount. Data preparation entails all activities which are employed in constructing data sets which shall be employed in the modeling phase. These shall incorporate data cleaning, combination of data attained from several sources then transforming them to variables that are more useful. Text analytics and feature engineering shall be employed in deriving new structured variables. This will enrich a set of predictors thus enhancing the accuracy of the model to be used.
Modeling and evaluation- modeling is accomplished through the use first version of data set which has been prepared. Historical dataset where there interest is well known will be used in modeling appropriate design required for this project. A predictive otherwise descriptive model shall be developed by the use of analytic approach which has been described above. The modeling process would be kept as iterative as possible. On evaluation, data scientists will be assessing the quality of model. Multiple checkups will be done to establish if the model is addressing the outlined business problem in the case or not. Conducting evaluation will call for many diagnostic measures on computing together with other outputs like graphs and tables. All these would be set in testing the predictive model to be used in the process.
With all the above described stages, the problem of fraud through the use of credit cards shall be curbed. Thus, it will lead to the reduction of problems facing financial and other businesses.
Works Cited
Angelov, Plamen P., Xiaowei Gu, and José C. Príncipe. “A generalized methodology for data analysis.” IEEE transactions on cybernetics 48.10 (2017): 2981-2993.
Schroeder, Greyce N., et al. “Digital twin data modeling with automationml and a communication methodology for data exchange.” IFAC-PapersOnLine 49.30 (2016): 12-17.