Climbing Mount Rainier- From a Machine Learning Point of View
Almost a year ago, I had the privilege of climbing one of the toughest mountains in the North American Cascade ranges — Mt RAINIER.
Mt Rainier is a 15000 ft heavily glaciated active volcano sitting just off Seattle, Washington. Every year hundreds of people attempt to climb the mountain and only one third succeed the Summit climb, mainly because of the rough terrain, unexpected weather changes en route the climb and depending on the fitness levels of the climbers. According to the NPS , there were about 10,940 people attempted the climb of which 5,250 were successful in reaching the rim of the crater over the past 10 years
Our team of 12 including 4 guides was a mixed group of people. A couple who have already attempted climbing Mt. Rainier before, three others who had experience with Mountaineering and the rest two including myself complete novice in Alpine climbing, but with some prior training. The entire team enjoyed a successful summit climb. We were noted as one of the strongest teams the guides worked with.
This article is an attempt to describe how Machine Learning Algorithms can be used to predict the success rate of a summit climb on a given day and through a given route, using historical climbing statistics and the weather information gathered over a period of time
The analysis was built on the data captured for the years 2014 and 2015. It would have been really useful if i had the data for the rest of the years.
Note that this ML Analysis and the Models does not consider the fitness levels of the climbers in assessing the rate of success but purely relies on the weather and the route which might not be a perfect model of assessment.
Lets look at some details from Exploratory Data Analysis:
The graph below shows the Number of Attempts and the Number of Successes in reaching the summit against the various months of the year .
The analysis shows that the peak period of climbing is during June, July and August.
The climbing season starts during the 2nd week of May and ends by September. I happen to climb during the 3rd week of May and during this period though the success rates are low due to the fact that the Summer is yet to start, the climbing is rather less harder due to the solid snow which gives a good grip on every step and the route could be shortened by climbing through the In-graham Glacier which would not be melted until the end of May. But watch out for those crevasses.
The graph below shows the various routes and the popularity.
As you see the Route through the Disappointment cleaver is the most sort out route. Its rather less harder to climb than through the other routes.
Now we know that routes and the duration when you climb do determine the success of a summit climb, it would be interesting to see how the weather parameters play a role on a given day.
The above picture is a heat map that describes the correlation of variables within the data set.
The battery voltage and the temperate have a negative correlation. The temperate and the solar radiation have a positive correlation. Humidity and temperature also have a negative correlation. In a way the Column Succeeded seems to be negatively correlated with all the variables except the Attempted and the Temperature.
Here is another way of looking at the correlation between the variables
With the analysis of the data here are few more steps done that helps in prediction
- Dimensionality reduction on highly co-linear variables and also using the feature importance
- Data Imputation for missing values and outliers
And finally applying the data to the models: I have used the below models to predict the success of the climb.
- Logistic Regression
- Support Vector Regressor
- Decision Tree and Random Forest
- Ensemble techniques with Bagging and Boosting
- Random Forest with Hyper Parameter tuning and Grid Search
The actual model code and the detailed steps are here — Kaggle- https://www.kaggle.com/codersree/regression-models-to-predict-success-rate
The model though could predict the success rate, the fitness condition of the climber plays a major role and
In reality its just pure will & grace that determines the success of the climb🤞
