Action Research is research methodology mainly used in education, it dissolve the barrier between the researcher and participants, in other word, the research actively involve in the situation and also conducting the research.. The process of action requires a cycle of steps to improve practice and taking action at the same time.
“Action research is a disciplined process of inquiry conducted by and for those taking the action. The primary reason for engaging in action research is to assist the actor (In IT is developer, actor, software) in improving and or refining his actions.” (Sagor, 2000)
The basic step of action research involves different phase:
- Planning phase
- Action phase
- Analysis phase
- Conclusion
First define a specific research question which can be tested. Then we can do some literature review to know more about the topic. After the know enough of the question, we can design how the research performs in a ethical way and set up the research proposal and setup deadline in each step of the research.
Second, in the action phase, it requires a cycle of experiment and data collection.
The purpose of the The data we collect can be qualitative data such as observation or interview. Or quantitative data like rubric data or survey.
Third, we need to organize the data we collected and perform data analysis. The proper way to organize data can be using chart, or graph to find any specific trend. Using static model to summarize or describe a collection
Finally, making the conclusion. Shar the research with the world by publishing blog, research paper or write up a book. This may lead up another new research question for other researcher and even create a new research area.
In IT field, we found some example adopted the similar method. example are reinforcement learning and software development.
Three are main Approaches in Machine Learning
- Supervised Learning
Supervised Learning:
Use labeled data to train the model
The machine knew the feature of the object and the label associated with the features
Machine base on the feature to predict the label
Unsupervised Learning
• Use unlabeled data
• Machine find some underlying structure to a dataset
• Aim to form groups depicting a sense of similarity.
To compare the difference between Supervised Learning and Unsupervised Learning, there are several areas we can focus on:
Method
Supervised Learning: Input variables and output variables will be given.
Un-Supervised Learning :Only input data will be given
Goal
Supervised learning goal is to determine the function so well that when new input data set given, can predict the output.
The unsupervised learning goal is to model the hidden patterns or underlying structure in the given input data in order to learn about the data.
Reinforcement Learning
It is a reward-based learning which works on the principle of feedback:
Figure above is an example of reinforce learning diagram.
First provide Environment for the agent(machine), then the agent gives out result. We give a negative feedback to the Agent for them to learn from the feedback and able to classify correctly next time.
In software development, normally we start with designing and gathering the requirements for the whole applications, then slice the project into functional portions that progress through the waterfall steps like the figure show below.
Albert's blog 