Our cutting-edge technologies and industry experience help businesses to use analytics insights to become smarter, more productive and more competitive.
Background
Automobile insurance fraud is a global problem. According to the General Insurance Association of Singapore, motor insurance fraud involves making false or exaggerated claims involving property damage or personal injuries as a result of an accident. The Coalition Against Insurance Fraud, a United States alliance working to enact anti-fraud legislation, puts the cost of fraudulent claims around US$80 billion annually. Handling fraud manually is costly for insurance companies. Insurers are intensifying efforts to educate the public on insurance fraud and using data analytics to track fraudulent and inflated claims.
Methodology
We developed a framework to assist insurance companies on the process of detecting fraud. This included identifying significant variables that contribute to fraud, such as customer profiles and claim types.
Results
We identified key variables for fraud detection such as demographics of claimants, claim characteristics, as well as policy and vehicle types. We then formulated a framework to detect fraud, including guidelines to alert insurance companies on suspicious records and possible fraudulent claims.
Impact
Profits for the insurance company significantly increased and investigation costs were reduced. Greater operational efficiencies also reduced business costs and saved time.
Background
Infections are the most common cause of death in infants less than four weeks old. Treatment of a neonate with signs of sepsis is required immediately, before the causative organism is known. In the developed world, neonatal severity scores have been created to estimate the risk of a neonate having a poor survival outcome. These scores rely on biochemical and haematological parameters, but these are often unavailable in the developing world.
Methodology
Using machine learning techniques, we built neonatal patient risk models to derive a mortality severity score for neonates in developing countries. The neonate mortality risk score was based on clinical signs that predict the likelihood of death.
Results
Our model accuracy was 85%, and included important factors that contribute to the neonatal mortality risk score, such as birth weight, temperature, heart rate and seizure.
Impact
We have shown that neonatal mortality risk can be reduced or prevented by scoring each neonatal at birth and then determining the level of care required to prevent death. Our neonatal mortality risk models have the potential to contribute to child health epidemiology, public health planning, monitoring and evaluation. Hospitals and healthcare authorities can also use the risk model to proactively identify high risk neonates, and the appropriate treatments and care required. Data analytics can therefore help healthcare providers make more informed decisions, open up possibilities in diagnoses and treatments, enable better care protocols and provide better insights into patient engagement issues. This will save more lives.
Background
Our clinical trial Phase 2 study aimed to measure the quality of life of patients with hepatocellular carcinoma (HCC), a primary malignancy of the liver. We also monitored different aspects of patients’ quality of life, to determine how these parameters changed over time.
Methodology
63 patients with advanced HCC were treated. The treatment benefits were evaluated through monthly assessments of baseline changes in patients’ quality of life.
Results
Convergent validity was supported by 52 of 57 hypothesised correlations. Predictive validity was supported by associations between survival and four aspects of quality of life. Patients’ direct assessments of treatment benefits for common symptoms were moderately associated with changes in their baseline scores.
Impact
We found that it is feasible to use the ‘quality of life’ forms, namely, the Functional Assessment of Cancer Therapy-Hepatobiliary Questionnaire (FACT-Hep) and the Patient Disease and Treatment Assessment Form (Patient DATA Form) in clinical trials of advanced HCC. Data analytics can therefore help healthcare providers make more informed decisions, enable better care protocols and provide better insights into patient engagement issues.
Background
To cater to the evolving needs of today’s discerning travellers, hotels will need to innovate and transform to stay ahead of the competition. Meeting customer expectations is key to customer loyalty. Increasingly, hotel operators are turning to advanced analytics solutions for deeper customer insights to enhance guest experiences.
Methodology
We performed a data exploratory analysis to evaluate hotel reviews and ratings. A factor analysis was also performed to identify key drivers of customer satisfaction. The scope of our analysis included: analysing customer ratings of hotel services; perceived differences between service levels offered by different hotels; identifying hotels that deliver higher value on factors that contribute to customer satisfaction; and analysing the competitive landscape of hotels.
Results
Perceptual mapping yielded insights into the competitive landscape of the hotel industry, while text analytics enabled us to identify positive and negative reviews as well as insights to improve customer satisfaction. We found that the top three attributes driving customer ratings are room quality and cleanliness; sleep quality; and service level.
Impact
During guests’ customer journeys, massive amounts of data are generated. With a Big Data strategy, hoteliers can gain deep customer insights which can create memorable experiences for their guests, optimise revenue, boost occupancy and improve service. Utilising our findings, staff were trained to ensure that room fittings, fixtures and cleanliness, as well as mattress quality / firmness met customers’ expectations. A dashboard was designed to monitor text analytics results, which provided a guide on timely and cost-effective issues management.
Background
As an efficient and low-cost mode of transportation, shipping is the backbone of global trade and the most carbon-efficient form of transporting goods. However, there is a need to further improve the fuel efficiency and carbon footprint of shipping vessels. The environmental impact of shipping includes greenhouse gas emissions, airborne emissions like sulphur oxides (SOx), nitrogen oxides (NOx) and ozone depleting substances.
Methodology
We established an emission control area, which covered shipping lanes along the Singapore Strait and the jurisdictional area of the Singapore port. Data was collected from ships operating in this area, with an Automatic Identification System. The maximum number of ships per day was determined as a baseline for emissions calculations. We then assessed emissions impact factors by quantifying CO2 emissions by area, and the quantity, type and flag country of shipping vessels.
Result
We found that Liquid Bulk tankers (i.e. merchant vessels designed to transport liquids or gases in bulk) are the highest contributors (33%) of CO2 emissions. Tugs (i.e. boats or ships that manoeuvre vessels by pushing or towing them) emit 28 times more CO2, compared to Solid Bulk ships, per gross tonnage of volume. We also found that a 10% reduction in the operational speed and a 10% increase in gross tonnage of Liquid Bulk and container ships can help to maintain CO2 emission levels for about two years, based on a 2% yearly projected growth rate of shipping volume / traffic.
Impact
Our findings will help the maritime industry to formulate more energy-efficient technical, operational and economic measures. This supports Singapore’s stepped up efforts to reduce the environmental impact of shipping and related activities, ahead of the 2020 target that the International Maritime Organisation has set for a 0.5% global cap on sulphur emissions from marine fuels, as well as green initiatives rolled out by the Maritime and Port Authority, including a scheme to adopt cleaner marine fuels.