Please enable JavaScript.
Coggle requires JavaScript to display documents.
Chapter 8: Trend in Data Technology - Coggle Diagram
Chapter 8: Trend in Data Technology
Data Discovery
Involves three main stages: data prep, visual analysis, and guided advanced analytics.
Demands skills in understanding data relationships, modeling, and employing analytics tools for uncovering insights.
Business user-focused method for spotting patterns and anomalies through visual exploration or guided analytics.
Big Data, Data Discovery, and Data Science
Data Science
Benefits
Complexity of analysis
Potential impact
Range of tools
Smart algorithm
Drawbacks
Slow and complex
narrow focus of analysis
Difficult to implement
Data Discovery
Benefits
ease of use
Agility and flexibility
time-to-results
Installed user base
Drawbacks
Limited depth of information exploration
Low complexity of analysis
Big Data
Benefits
Volume, velocity, or variety of data
Potential business impact
Drawbacks
Difficult to implement
Potentially expensive
Lack of skills available
Deep Learning
Deep learning is a part of AI specializing in creating complex neural network models for making accurate decisions based on data.
Ideal for handling complex data and large datasets.
technologies use deep learning
Facebook employs deep learning to analyze text in online discussions.
Google, Baidu, and Microsoft utilize deep learning for image search and machine translation.
Most modern smartphones incorporate deep learning for tasks like speech recognition and face detection in photos.
In healthcare, deep learning is applied to interpret medical images like X-rays and MRI scans for diagnosing health issues.
In self-driving cars, deep learning aids in tasks such as localization, motion planning, environment perception, and tracking driver status.
Machine learning, and Deep learning
Machine Learning
creates algorithms for computers to learn functions from datasets.
Deep Learning
extracts patterns from large datasets to make accurate decisions based on complex inputs.
Machine Learning
A machine learning algorithm searches for the best function among many possible ones to explain relationships in a dataset.
Understanding how a function is learned involves looking at sample inputs and the corresponding outputs of an unknown function.
Key ingredients
Data
Historical examples used for learning.
Functions
A collection of potential functions searched by the algorithm to find the best match with the data.
Fitness Measure
Used to assess how well each candidate function matches the data.
Types of machine learning
Unsupervised Machine Learning
Unsupervised machine learning is utilized for clustering data, like customer segmentation.
The algorithm aims to identify functions mapping similar examples into clusters.
Clusters group together examples that are more similar to each other than to examples in other clusters.
Fitness functions reward candidate functions promoting higher similarity within clusters and diversity between clusters.
Reinforcement Learning
Reinforcement learning is crucial for tasks like robot control and game playing.
Agents learn policies on how to act in environments to receive rewards.
Agents aim to map current observations and internal states to appropriate actions.
Deep learning is applicable to supervised, unsupervised, and reinforcement learning scenarios.
Supervised Machine Learning
The algorithm evaluates the fitness of candidate functions by measuring the difference, or error, between outputs and targets.
These target values help the algorithm by allowing it to compare function outputs with the specified targets.
In supervised machine learning, each dataset example is labeled with an expected output value.
Fitness evaluations guide the algorithm in searching for the best function.
Benefits from business perspective for using machine learning
Reduce data entry errors:
Implement spell checkers and pattern matching.
Minimize inaccuracies in data entry processes.
Enhance data quality and reliability.
Improve financial rule and modeling precision:
Increase accuracy and efficiency of financial processes.
Reduce errors and mitigate risks.
Enhance portfolio management and fraud detection.
Forecast medical and employee downtime:
Analyze trends in employee absences.
Optimize workforce management.
Ensure uninterrupted business operations.
Foresee maintenance needs:
Predict equipment maintenance requirements.
Optimize maintenance schedules.
Reduce downtime and maintenance costs.
Predict future sales accurately:
Monitor sales trends in real-time.
Adjust sales channels accordingly.
Ensure adequate product availability.
Augment customer interaction and improve satisfaction:
Enhance customer engagement and satisfaction.
Improve overall customer experience.
Analyze customer data for personalized solutions.
Simplify product marketing:
Determine optimal timing for sales campaigns.
Ensure effective utilization of resources.
Analyze customer preferences.
Machine Learning Limitations
Machines can't explain themselves:
Hidden functionality in machine learning models.
Deep learning solutions contain multiple hidden layers.
Understanding complex models becomes challenging.
Bias makes the results less usable:
Algorithms can't detect biases in data.
Data is treated as unbiased and truthful.
Analysis from biased data can be unreliable.
Labelling data is tedious and error-prone:
Supervised learning requires data labelling.
Labelling large datasets is time-consuming.
Each data point needs individual labelling.
Machine learning solutions can't cooperate:
Lack of collaboration capability in machines.
Human collaboration leads to exponential knowledge growth.
Machines remain isolated solutions without the ability to generalize knowledge or contribute to comprehensive solutions.
Massive amounts of training data are needed:
Machine learning relies on extensive data for training.
Problem complexity increases data requirements.
Some domains lack sufficient data or processing power.