- Business Understanding :
The main objective of this data is to do time series analysis and predict the number of fails. Here the transfer of data from the sender and receiver in the field of Communication is compared to raven and the sender and receiver are compared to the characters of Game of Thrones. Here there is a need of domain knowledge in the field of communication to understand the data to perform analysis on the data.
- Data Understanding:
The given data is a comparison of the olden day communication system using ravens and modern day mobile communication system. For a raven to reach from the sender to receiver it has to face many obstacles in the way before reaching the destination, similarly the communication signals also faces many obstacles such as packet loss, connectivity issues. Here there are lot of parameters to consider to find whether the communication will happen successfully from the sender to receiver.
- Data Preparation:
The first step for doing analysis is to find the granularity of the data. Each and every column of the data has to be analysed and understood before utilizing the data to build the model. Exploratory analysis of the data was done to find which ravens have the highest success rate and lowest failure rate of the ravens. As every row in the data is a failure of the raven, the success and failure rate is calculated under the assumption that if all the column values (factors) are 0, then none of the factors affected the communication .To find the family with the least and maximum raven fails first the count of failures and successes are found and from that each family count is derived.
Below is the graph for the least successful and most successful ravens :
For performing the Time series analysis using ARIMA the data was converted to uni-variate model where for each date the number of failures is counted,so that while building the predicting model for the next 4 days the number of failures can be calculated.
Using ARIMA model the likelihood value obtained is-141.64 for the value of and the forecast of the failure count for the next 4 days are
Day 1 :9481
Day 2: 10182
Day 3: 10021
Day 4: 9884
Hey guys, great work!
If I understand correctly, you’ve used ARIMA on the total failure (from all columns), correct? It will be nice to detail it a bit more in the article.
Also, which other forecasting methods would you consider using? Which of the features are the most prune for causing raven-delivery failure? or in other words – how would you reason your prediction results?
Consider writing which further actions would you recommend the family to take, in order to maximize delivery (which failures should they focus on fixing)