Our case menu allows you to select the cases that would most relate to your workforce as we build your customized Analytics Edge for Leaders platform

Our Case Menu

Strategy

Case
Describe
Predict
Prescribe
Supporting Tools
Quantify Value Add
Strategy

Capital Intensive ⋅ Industrials

  • Construct a table of decision outcomes
  • Maximize value by choosing to invest in, maintain, or divest 50 business units at a conglomerate using integer optimization
  • Establish an objective function, decision variables, and constraints
  • Test the module for robustness
  • Determine maximum value and compare to a defined baseline
Modeling COVID-19 Progression and Policy Effects

Public Sector ⋅ Healthcare

  • Defining a source of truth
  • Predict COVID-19 hospitalizations and mortalities
  • Dynamic modelling: estimate trends based on policy decisions
  • Sparse data conditions
    • Meta analysis
    • International multi-center study
  • 'All models are wrong. Some are useful'
  • Compare models by assessing accuracy (predicted vs. actual case count)

Marketing & Sales

Case
Describe
Predict
Prescribe
Supporting Tools
Quantify Value Add
Movie Recommendations

Technology ⋅ On-Demand Streaming

  • Choose a filtering method
  • Use hierarchical clustering to group movie genres
  • Recommend list of movies based on preferred genre
  • Develop a decision support tool
  • Leverage human judgement to choose the number of clustering groups
Blue Bikes

Commerce ⋅ Transportation

  • Review hierarchal clustering
  • Review K-Means clustering
  • Leverage optimal classification trees to build an interpretable clustering model
  • Inspect a dendrogram
  • Apply human judgement to sense check cluster output
Sentiment Mining with Twitter

Commerce ⋅ Telecommunications

  • Use natural language processing to convert tweets to numeric matrix format
  • Use classification and regression trees (CART), Random Forest, and logistic regression to predict Tweet Sentiment in order to automate tweet labelling
  • Develop a confusion matrix
  • Balance confusion matrix accuracy and interpretability
Real Estate Pricing

Capital Intensive ⋅ Real Estate

  • Explore 1-D and 2-D graphing methods
  • Conduct geospatial exploratory data analysis
  • Predict home prices
  • Inspect a data dictionary
  • Use R-squared to identify linear regression model with best fit

Operations

Case
Describe
Predict
Prescribe
Supporting Tools
Quantify Value Add
Air Force Tanker Routing

Capital Intensive ⋅ Aerospace & Defense

  • Create a set of data tables based on the problem statement
  • Determine the optimal number of tankers, where to bed them down, and where they should fly based on an objective function, a number of constraints, and decision variables
  • Translate plain language to a mathematical expression to computer code
  • Compare the output of an optimization engine to the status quo or baseline
  • Verify the answer is robust by determining the second-best answer and identifying a set of second order questions
Framingham Heart Study

Public Sector ⋅ Healthcare

  • Univariate data inspection
  • Create a naive baseline
  • Use logistic regression to predict coronary heart disease
  • Create a training/test set
  • Develop a confusion matrix
  • Develop decision support tool
  • Assess bias by distinct cohort studies
  • Use AUC to determine model effectiveness
Supply Chain Optimization

Capital Intensive ⋅ Mining

  • Build a mixed integer universal supply chain model
  • Create a process map and define decisions
  • Define derivative models from the universal supply chain model
  • Establish financial metrics
  • Define value to the organization
Interventional Radiology Scheduling

Public Sector ⋅ Healthcare

  • Create a set of data tables based on the problem statement
  • Determine the optimal schedule for a team of interventional radiologists that maximizes revenue in a fair capacity
  • Define the concept of fairness and create a fairness metric
  • Design hyperparameters to modify the algorithm to trade-off total revenue and physician accommodation
  • Compare the output of an optimization engine to the status quo or baseline
  • Quantify the improvement in fairness
Automated Market Research at Tinker Air Force Base

Capital Intensive ⋅ Aerospace & Defense

  • Cluster companies into topic groups
  • Conduct labeling leveraging human judgement to train a semi-supervised learning algorithm
  • Identify companies to autonomously contact via email
  • Procure a dataset
  • Leverage human judgement to inspect the output of a semi-supervised training algorithm
Immigration Enforcement

Public Sector ⋅ Legal

  • Inspect a dataset for trends
  • Develop an optimal classification tree (OCT) predictive model
  • Create a "difference tree" to determine if a policy is being followed
  • Identify data drift, also called regime change
  • Use AUC to determine model accuracy
Market Model

Capital Intensive ⋅ Mining

  • Predict market demand and price for fertilizer products
  • Differentiate readily available and knowable data from not readily available data
  • Identify data sources to support a model driven modeling approach
Interpretable Predictive Maintenance for Turbofans

Capital Intensive ⋅ Energy

  • Conduct univariate visual inspection of input data
  • Predict next day failures of turbofans (CART, XGBoost, Optimal Classification Trees)
  • Predict survival time distribution (Optimal Survival Trees)
  • Create a training and test dataset
  • Training and testing set AUC
  • Harrell's c-statistic
Capacity Planning

Capital Intensive ⋅ Construction

  • Generate variable (stochastic) cost estimates from an original point estimate
  • Predict total costs from variable (stochastic) cost estimates
  • Build a deterministic optimization model
  • Build a robust optimization model
  • Visualize variable (stochastic) cost estimates
  • Visualize outcomes derived from variable inputs
  • Compare the outcome of the robust optimization model to the deterministic baseline model.
Hard Drive Failure

Commerce ⋅ Computer Hardware

  • Determine the baseline failure rate of hard drives
  • Inspect standard Internet of Things (IoT) metrics
  • Transform IoT metrics to normalized values
  • Predict long-term failure of a hard drive with optimal classification trees
  • Build long-term and short-term survival curves to guide investment decisions and schedule maintenance
  • Build models in a data-light environment
  • Build survival curves
  • Use area under the curve (AUC), accuracy, sensitivity, false alarm rate, and precision to determine model effectiveness
Handwriting Recognition

Technology ⋅ Computer Vision

  • Find the lines of handwritten text in a document with neural networks and image processing
  • Predict the probability of character presence in a line of text using neural networks
  • Arrange the predicted characters into words that form meaningful sentences using optimization
  • Breaking up the project into building blocks and evaluating improvement
  • Introduce performance metrics tailored to your work
Default Risk

Financial Services ⋅ Banking

  • Univariate data inspection
  • Predict loan default with OCT and XGBoost
  • Sense check: evaluate individual cases
  • Compare models by assessing AUC and interpretability
Capital Expenditure

Capital Intensive ⋅ Mining

  • Estimate a notional future operating model
  • Build a mixed integer optimization model to support capital expenditure decisions
  • Create a process map and define decisions
  • Define a multi-faceted objective function consisting of both existing assets and notional assets

Organization

Case
Describe
Predict
Prescribe
Supporting Tools
Quantify Value Add
Talent Management

Entertainment ⋅ Baseball

  • Use linear regression to predict
    • season wins
    • runs scored
    • runs against
  • Conduct feature engineering
  • Build a decision support tool
  • Use adjusted R-squared to identify linear regression model with best fit

Risk

Case
Describe
Predict
Prescribe
Supporting Tools
Quantify Value Add
Congenital Heart Surgery

Public Sector ⋅ Healthcare

  • Orient to the current state: the European Congenital Heart Surgery database
  • Predict surgical outcomes: mortality, mechanical ventilator support time (MVST), and prolonged length of stay (los)
  • Leverage a clinical decision support tool to support decisions under varying scenarios
  • Benchmark a hospital against ~180 European hospitals
  • Inspect analysis output
  • Identify areas of opportunity for quality improvement
D2Hawkeye

Public Sector ⋅ Healthcare

  • Conduct exploratory visual analysis of healthcare data
  • Predict healthcare spend with classification trees
  • Infuse subject matter expert guidance into the modeling process with a penalty matrix
  • Leverage cross validation
  • Use penalty error to compare baseline performance to CART performance
COVID-19 Infection and Mortality Calculator

Public Sector ⋅ Healthcare

  • Hypothesis generation
  • Predict COVID-19 infection and mortality
  • Sparse data conditions
    • Meta analysis
    • International multi-center study
  • Review a clinical decision support tool
  • AUC vs. interpretability
  • SHAP plots
Opioid Prescription

Public Sector ⋅ Healthcare

  • Provide supervisory labels using external datasets
  • Use cluster analysis to generate insights
  • Create supervised and semi-supervised models to predict likelihood of illegal drug diversion
  • Create fact bases to justify further research or investigation activity.
  • Test for both accuracy and robustness
Surgical Risk

Public Sector ⋅ Healthcare

  • Orient to the existing database: The American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP)
  • Predict surgical outcomes (mortality and eighteen specific co-morbidities) with an optimal classification tree
  • Evaluate algorithms by inspecting derivation cohort, validation cohort, and entire cohort
  • Evaluate model AUC
Insuring Capital Equipment

Capital Intensive ⋅ Aerospace & Defense

  • Conduct exploratory visual analysis of aircraft fleet data
  • Predict annual airline costs
  • Use simulation to determine which of four insurance policies to choose
  • Visualization of simulation output
  • Use expected cost and standard deviation to guide decision
Structuring Data to Predict Stroke

Public Sector ⋅ Healthcare

  • Assess the capabilities and limitations of natural language processing techniques:
    • bag of words
    • term frequency-inverse document frequency (TF-IDF)
    • global vectors for word representation (GloVe)
  • Predict stroke presence, location, and acuity
  • Conduct text preprocessing to start the process
  • Compare and analyze the output of seven prediction algorithms using AUC
Jury Selection

Public Sector ⋅ Legal

  • Leverage optimal feature selection to identify variables of importance
  • Leverage an optimal classification tree without key variables to test for bias
  • Conduct subgroup analysis to test for disparity
  • Leverage AUC to determine model accuracy

© 2024 Dynamic Ideas LLC. All rights reserved.