Subscribe to DSC Newsletter

Rule of thumb: Which AI / ML algorithms to apply to business problems

 

How to know which AI/ ML algorithm to apply to which business problem?

This is a common question

I found a good reference for it – Executive’s guide to AI by Mc Kinsey

I summarize the insights below

 

Firstly, there are three broad categories of algorithms:

  • Supervised learning: You know how to classify the input data and the type of behavior you want to predict, but you need the algorithm to calculate it for you on new data
  • Unsupervised learning: You do not know how to classify the data, and you want the algorithm to find patterns and classify the data for you
  • Reinforcement learning: An algorithm which learns by trial and error by interacting with the environment. You use it when you don’t have a lot of training data; you cannot clearly define the ideal end state; or the only way to learn about the environment is to interact with it

 

So, let us consider which algorithms can apply to business problems

 

Customer services and supply chain

  • Understand product-sales drivers such as competition prices, distribution, advertisement, etc linear regression
  • Optimize price points and estimate product-price elasticities linear regression
  • Classify customers based on how likely they are to repay a loan logistic regression
  • Predict client churn Linear/quadratic discriminant analysis
  • Predict a sales lead’s likelihood of closing Linear/quadratic discriminant analysis
  • Detect a company logo in social media to better understand joint marketing opportunities (eg, pairing of brands in one product): Convolutional neural networks
  • Understand customer brand perception and usage through images : Convolutional neural networks
  • Detect defective products on a production line through images: Convolutional neural networks
  • Power chatbots that can address more nuanced customer needs and inquiries Recurrent neural networks
  • Assess the likelihood that a credit-card transaction is fraudulent Recurrent neural networks
  • Predict call volume in call centers for staffing decisions Random forest
  • Detect fraudulent activity in credit-card transactions. (Achieves lower accuracy than deep learning) AdaBoost
  • Provide a decision framework for hiring new employees Decision tree
  • Understand product attributes that make a product most likely to be purchased Decision tree
  • Forecast product demand and inventory levels Gradient-boosting trees
  • Predict the price of cars based on their characteristics (eg, age and mileage) Gradient-boosting trees
  • Recommend what movies consumers should view based on preferences of other customers with similar attributes Recommender system
  • Recommend news articles a reader might want to read based on the article she or he is reading Recommender system
  • Stock and pick inventory using robots Reinforcement learning
  • Cluster loyalty-card customers into progressively more microsegmented groups Hierarchical clustering
  • Inform product usage/development by grouping customers mentioning keywords in social-media data Hierarchical clustering
  • Segment customers to better assign marketing campaigns using less-distinct customer characteristics (eg, product preferences) Gaussian mixture model
  • Segment employees based on likelihood of attrition Gaussian mixture model
  • Segment customers into groups by distinct charateristics (eg, age group)— for instance, to better assign marketing campaigns or prevent churn k means clustering
  • Analyze sentiment to assess product perception in the market Naive Bayes
  • Create classifiers to filter spam emails Naive Bayes
  • Predict whether registered users will be willing or not to pay a particular price for a product Simple neural network
  • Predict how likely someone is to click on an online ad Support vector machine

 

 

Healthcare

  • Predict if a skin lesion is benign or malignant based on its characteristics (size, shape, color, etc) logistic regression
  • Diagnose health diseases from medical scans : Convolutional neural networks
  • Predict the probability that a patient joins a healthcare program Simple neural network
  • Predict how many patients a hospital will need to serve in a time period Support vector machine

 

 

 

Trading

  • Optimize the trading strategy for an options-trading portfolio Reinforcement learning
  • Optimize pricing in real time for an online auction of a product with limited supply Reinforcement learning
  • Generate analyst reports for securities traders Recurrent neural networks

 

 

Other

  • Simple, low-cost way to classify images (eg, recognize land usage from satellite images for climate-change models). Achieves lower accuracy than deep learning AdaBoost
  • Optimize the driving behavior of self-driving cars Reinforcement learning
  • Balance the load of electricity grids in varying demand cycles Reinforcement learning
  • Predict power usage in an electrical- distribution grid Random forest
  • Provide language translation Recurrent neural networks
  • Track visual changes to an area after a disaster to assess potential damage claims (in conjunction with CNNs) Recurrent neural networks
  • Generate captions for images Recurrent neural networks

 

Notes

Linear/quadratic discriminant analysis: Upgrades a logistic regression to deal with nonlinear problems—those in which changes to the value of input variables do not result in proportional changes to the output variables.

 

Gaussian mixture model: A generalization of k-means clustering that provides more flexibility in the size and shape of groups (clusters)

 

Image source: Executive’s guide to AI by Mc Kinsey

 

Views: 9612

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Comment by Sreedhar sree on March 14, 2020 at 6:57pm

Thanks Ajit for sharing such an infirmative post .. I'm working on a POC and problem statement for that is

Problem statement : IVR systems contained call rate is decreasing from last 2years and number of calls being transferred to agents are increasing because of which we are spending more money for resources .. we are trying to find out what is the reason for this decrease in contained call rate using ML ... In this case which algorithm we use to find the reason for reduction in contained call rate ..

Comment by Sreedhar sree on March 14, 2020 at 6:56pm

Videos

  • Add Videos
  • View All

© 2020   Data Science Central ®   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service