Article Preview
Top1. Introduction
Organizations rely on processing and acquiring meaningful information from raw data and challenge arises when amount of data exceeds the nominal level that could be processed by a human expert (Forbes, 2017). This limitation led to the development of an interesting domain in Computer Science and Engineering namely business intelligence (BI) (Teixeira, Afonso, Oliveira et al., 2015). Data warehousing optimizes the storage and retrieval of voluminous data, with the support of powerful business analysis tools, these data warehouses extends its functionality to a wider domain. The heterogeneous data are classified and clustered for easier processing by online analytical processing tools to aid in efficient decision-making processes. Yet in medical industries, types of tests and reports possess a great diversity which makes the condition more challenging. The tools implied will also face the difficulty of extracting the required information from multiple data fields. Possibilities of misguidance and misjudgments always yield potential harm to the patients’ health and reputation of the medical institutions. The decision support systems cannot always rely on training data and a common pattern analysis (Peng, 2005) as the observable characteristics are represented to be referral. Proofs from the reports may seem similar to a pattern tested during investigative stages but response of the particular patient to a drug may differ from representative diagnosis by a great extent. This case is best understood with the probability of diabetes when it possesses a family history. The children of hereditary patients affected with diabetes need to be tested in earlier years when compared to children of healthier patients. The same scenario applies to patients of breast cancer. Hence this system should be able to decide the age and interval of mandated tests based on family history and other risk factors (Richmond, 2008). Even the diagnosis prescribed by qualified doctors are wrong sometimes and account to nearly 10% of mortality rates, which questions the development of a business intelligent architecture for decision support on prescription and diagnosis (Bennett, 2013).
This proposal intends to develop a medically intelligent tool backed by sufficient learning algorithms to demonstrate and justify the need of a clinical decision support system (CDSS) (Berner, 2007) and diagnosis decision support system (Çomak, 2007) in hospitals. Patients’ information is maintained electronically as electronic health records (EHR) (Berner, 2007). These records are constituted with the details of patients’ details apart from name and address including their demographics, age, allergic reactions, previous medications, progress and regress of previous treatments, dosage and so on. This abundant information provides an additional opportunity to understand the physique and mental stages of patients. These details will assist in a customized diagnosis plan by eliminating the odds. The intelligent business model is already trained with similar patients, their treatment plans and counter actions (Shortliffe, 2013). Yet this tested result may not be the right plan for a new patient with similar characteristics. The objective is to derive the distinguishing factors present in the current patient and compare with the treatment plans which are already tested to be efficient. This step of the model is to preserve the computation cost and time to ensure an efficient product for the clients. Comparative study over the existing metrics will delay the overall processing and result in poorly performing software solution. The adaptability of the system to new conditions and parameters will also be reduced when the number of patients is more than the testing environments. This is coined as the Perception Factor of the proposed algorithm which concentrates more on distinguishing parameters in a large patient health record.