Most of the organizations are heavily investing in advanced analytics and predictive modeling. This can be the motivation for why we have chosen this course. These investments are mainly due to three things. The first thing is the number of devices like a laptop, smartwatch which are connected to the internet and emitting data on a daily basis. Each of these devices is capturing data and sending it to the cloud. Second thing, the cost of data storing data now is less than what it used to cost in the 1970s. The third thing is that Machine learning significantly reduces the cost of computational data. The process of generating predictive models involves six stages:-
1. Problem Definition
The first step in predictive modeling is to find a problem you want to solve. It might sound very instinctive. But it gets ignored most of the time.
Hypothesis generation is the process of listing down all possible variables which might influence problem objectively. The quality of the model is heavily dependent on the hypothesis generation. Hypothesis generations should be free from personal bias.
Example of hypothesis generation: Think about what are the factors which will affect the price of the house. The price of a house is dependent on parking facility, locality, number of bedrooms, balcony. These are a certain hypothesis that you can think about.
3.Data Extraction / collection
After completing hypothesis generation, we will have knowledge of what you are looking for. We can collect data from different sources to either approve or disapprove your hypothesis.
Data exploration is an approach to the analysis of data, where data analyst uses visual exploration to gain insight into data. This process separates good analysts from bad analysts. A good analyst knows his/her data very well.
5. Build a model
First step in this is a selection of modeling techniques based on the defined goals(supervised, unsupervised) such as logistic regression, linear regression, decision tree. Next is to train a model and check the model performance.
6. Model Deployment / Implementation
Deploying a machine learning model means to integrate a machine learning model and integrate it into an existing production environment where it can take in an input
and return an output
Generating high-quality predictive models is a time-consuming activity because of the tunning process in finding the optimum model parameter and often required to redevelop, reuse model in the future. Thus it is important to follow standard methodologies and establish governance in model building.
Comentários