SESSION 11 - Machine learning for enterprises Flashcards
What is machine learning?
Machine learning transforms human processes into intelligent, automated processes, which allows enterprises to focus their resources toward higher-value activities (like customer retention & acquisition)
one of the must disruptive innovations and strong enabler of competitive advantage for businesses
can reduce cost by between 20-25%
can generate new revenues across products and services
can improve customer retention and acquisition
Whats the relationship between AI, machine learning and deep learning?
Artificial Intelligence (AI)= a broader domain than machine learning that includes speech and image recognition, natural-language processing (NLP), and object manipulation
–>AI is necessary for machine learning
Machine learning= a subset of artificial intelligence that can learn patterns from data without the need to define them a priority
Deep learning= is a subset of machine learning, employs a com-plex structure with many nodes, hidden units, and learning algorithms, all of which influence training time and quality of learning
–>Deep learning is expected to be at the center of the machine-learning field
What are the three categories of machine learning algorithms?
Supervised Machine Learning
Unsupervised Machine Learning
Semi-supervised Machine Learning
What are the three main machine-learning applications used by enterprises?
Clustering= used to group sets of objects on the basis of their similarities in a multidimensional space
Classification= the process of identifying the category or class of an observation
Prediction= Machine learning is used to identify patterns in data and to predict future events
Between which two attributes exists a tradeoff?
-One of the challenges in choosing the best algorithm is managing the trade-off between accuracy and interpretability
–> Accuracy= measure of how well the algorithm will perform in practice
–>Interpretability= the ability to explain to users how a particular decision or response is made
–> trade-off between accuracy and interpretability arises for two reasons:
–>by adding more parameters, a model’s accuracy increases, but the outcomes become harder to understand and interpret
What are potential types of errors?
-Reducible errors= originate from the fact that the chosen model will generally not be a perfect estimate of the true function, and this inaccuracy will introduce some errors
-Bias error= refers to errors introduced through faulty assumptions about the nature of the function
-Variance error= the amount by which the estimate of the function learned from one training data set would change if a different training dataset were used, occurs because of sensitivity to small fluctuations in the training data set
-Irreducible errors= even if a machine-learning algorithm were able to capture perfectly the true function, some errors would be irreducible because of inaccurate data and missing predictor variables
What are challenges in deploying machine learning at enterprises?
- The ethical challenge:
-When collecting training data, a company must take care to comply with data privacy and protection rules - The storage of machine learning engineers
-it will take years to adequately meet the industry demand for machine-learning-educated jobseekers - The data quality challenge
-When data are more unstructured and collected from more sources, data quality tends to decline - The cost-benefit challenge
-Since machine learning is not a solution for all business problems, managers need to have a clear understanding of its value-generation mechanism
Whats the conclusion of the article?
As machine learning pervades, managers who learn early on about machine-learning tools and techniques can quickly identify opportunities and potential benefits, effectively communicate their potential to stakeholders, and bring competitive advantages to their enterprises