dsckgec / nasa-collision-detection Goto Github PK

Is your feature request related to a problem? Please describe.
There are 8840 values for hazardous objects and the remaining 81995 values for non-hazardous objects which clearly indicates a large imbalance in the dataset.

Describe the solution you'd like
So using the concept of upsampling(hazardous) or downsampling(non-hazardous) whichever solution produces the optimal results to be implemented.

Additional context
Using the from sklearn.utils import resample will produce the code.

The data set is biased . So the accuracy level is not good . Need help for cleaning the data and removing the bias. There are 8840 values for hazardous objects and the remaining 81995 values for non-hazardous objects which clearly indicates a large imbalance in the dataset.
To remove bias preferably sub and super sampling is needed.

Univariate statistics summarize only one variable at a time. Bivariate statistics compare two variables. Need help to do this for this data set . And show the output . And explain the importance of it.

Data visualization is not done . Need help For data visualization for better impact . Any kind of graphs or plotting is preferable. Plot histogram or density graph or Box Plot to visualize, "est_diameter_min", "est_diameter_max", "relative_velocity", "miss_distance" , "absolute_magnitude". of the data set.

To improve the EDA, we can add categorical variables based on the columns of absolute magnitude, relative velocity and estimated diameters

We can create two or three categories per variable (ex: in absolute magnitude, low, medium and high magnitude) and use a boxplot to identify tendencies in these categories based on miss distance.

Apply Random Forest Classifier for supervised machine learning for the Classification. In the output accuracy and confusion matrix must be printed. Show the accuracy is improved or not . A random forest is a meta estimator that fits a number of decision tree classifiers on various sub-samples of the dataset and uses averaging to improve the predictive accuracy and control over-fitting. Do the necessary plotting to tally between hazardous and non hazardous nearest earth objects.

Apply Logistic Regression for supervised machine learning for the prediction. In the output accuracy and error calculation must be printed. Show the accuracy is improved or not . Since the outcome is a probability, the dependent variable is bounded between 0 and 1.

We need proper documentation and organization of the code so that it can be easily understood by newcomers. Make sure that every level of code should have proper headings and sub-headings according to their work. . A well organized notebook of code is much needed.

Apply KNN algorithm for supervised machine learning for the prediction. In the output accuracy and confusion matrix must be printed. Show the accuracy is improved or not . Plotting should be done between hazardous and non hazardous earth's nearest object.