Improving the accuracy of a predictive model is difficult at times and people get stuck at times but this is real situation where real story begins.
A predictive model can be built in many ways their is no rule but if you follow the ways shared below will increase the accuracy of the predictive model.
1. Add more data.
Presence of more data always results in more accurate model and allows the data to speak for itself than relying on assumptions.we can always ask for more data in order to improve the accuracy.
2. Treat missing values.
Missing values can create a problem while building a predictive model.This is because we dont analyze the relationship with other variables correctly.
These missing values can be treated by different ways such as replacing the missing value with the mean or median of the data. Missing string data
can be replaced by the mode of given data outliers can be treated by removing them from data but it is not always good idea to remove data.
Changing the scale of variable from normal scale to scale between 0 to 1 is data normalization. some algorithms work really very well when data is distributed normally therefore e should remove the skewness of variable.we must normalize the whole data into same scale so as to improve the accuracy of the predictive model.
4. Feature Selection
It is the process of finding out the best subset of attributes of given data which can better explain relationship of independent variable with the target variable.
You can select subset of attributes based on different matrices such as domain knowledge,data visualization etc. it will surely help in making better predictive model.
5. Use Different and Multiple Algorithms
Always try to use different algorithms and check which algorithm gives a better result according to the data and you can also combine multiple algorithms to get more accurate model. hitting the right algorithm is a very important task. Always check the performance of model using different algorithms and hit the correct one.
Building a predictive model is difficult but these steps can help in making a better predictive model.
Data always speaks, all you need is to listen. 🙂