PART C
Introduction
This report aims to analyze the determinants of the demand for renting bikes. The data used in this report is collected from a database containing about 731 rentals. The data in the database has eight characteristics which are the most likely to be influencing the demand for bikes. The characteristics include the record number which acts as the index, the season which has four categories, the holiday which has two categories, and the weather situation which has four categories, the humidity, the wind speed, and the total number of rental bikes.
Data preparation, descriptive statistics, and presentations.
The table below shows the descriptive statistics for the data. Descriptive statistics containing the measures of central tendency and variation can only be calculated on variables with a continuous scale or the ratio data.
Table 1 Descriptive statistics for selected variables
The data used is of the whole population of rental bikes. The average normalized apparent temperature is about 0.4743 and a median value of 0.486733. The proximity of the mean to the median suggests the absence of outliers in the temperature. The standard deviation and variance are very low indicating that most values are close to the average value. For the normalized humidity divided to 100, the mean is 0.6279 while the median is 0.6267. For this variable, the median and mean are also in proximity. The standard deviation and variance in the humidity are very small indicating that the difference between mean and actual values is minimal. The wind speed divided by 67 has an average value of 0.1905 and a median value of 0.180975. The median and average values also have a small difference. The standard deviation and variance are relatively small indicating the proximity of the actual humidity values and the average value. The data is not skewed and is relatively peaked as indicated by the values of the kurtosis and skewness being between +1 and -1. There is an average of 4504.349 number of rental bikes is 4504.349 while the median count was 4548 bikes. The difference between the mean and median is substantial implying that the number of bikes available may have outliers.
The figures below show the histograms for the data.
Figure 1 Histogram for temperature
Figure 2Histogram for humidity
Figure 3 Histogram for wind speed
Figure 4 Histogram for the count of rental bikes
From the observation of the histograms above, all the variables show some extent of the presence of outliers as shown by the presence of values in the region below the lower quartile and upper quartiles.
The database also contains nominal data which can be analyzed using frequencies.
Figure 5 Histogram for the distribution of data in different seasons
The data was collected over the four seasons. The number of data points in all seasons was almost similar to the summer having the highest and autumn having the lowest.
Figure 6 Histogram for the distribution of data in weather conditions
According to the figure above, the Majority of the data collected on the bikes was collected on the days when the weather was holiday.
Figure 7Histogram for the weather situation
From the table above, most of the data on rented bikes were collected when the weather is clear, there are few clouds, or when it was partly cloudy. No number of entries were recorded when there was a combination of heavy rain, ice pallets, thunderstorms, mist, snow, and fog. As the weather conditions deteriorate, the number of people renting the bikes decreases as the weather deteriorates.
Analysis
This report aims to identify the factors that are likely to influence the demand for bikes for rent. The table below shows the correlation between the counts of bikes rented out other characteristics in the data set.
Table 2 Correlation analysis
Figure 8 humidity vs cnt
Figure 9 Temperature vs cnt
Figure 10 Wind speed vs cnt
The linearity in between the count and other attributes is described in the scatter plots above. There exists a linear inverse relationship between the humidity, wind speed, and the variable cnt. The negative linear relationship between the two variables implies that as one variable increases the other variable is likely to decrease. There exists a positive linear relationship between the temperature and the count of bicycles that rented out. As the temperatures increase, the number of bicycles rented out also increases.
Task translation and interpretation.
A multiple regression model can be developed from the data to determine the significance of the different factors on the demand for bikes.
Table 6 Regression table
The larger the value of the coefficient, the larger the significance of the variable in the study. The most significant variable is the temperature followed by the wind speed. The least significant factor is the season. The model used to forecast the demand for bicycles can be represented by the equation:
The table below shows the accuracy of the model.
Table 7 Regression analysis
From the table above, the model represents the data with an accuracy of about 51.76% of the data. This model is good for theoretical forecasts that can be used in planning.
Evaluation.
In the analysis, the relationship between the factors affecting demand of the bicycles and the count of bicycles rented is determined. From the relationship, it can be concluded that the demand for rented bikes is experienced during the summer. The summer has the highest temperature. The strong positive relationship between demand for rented bikes and temperature supports the idea that demand is highest in the summer and this is mainly due to the high temperatures. The temperatures also determine the weather conditions. The majority of the people prefer the days the weather is holiday and relatively clear. From the regression model represented by an equation in analysis, the demand for bikes can be predicted. Predictions help in making decisions for the owners of the bikes. From predictions, it is possible to predict a likely increase or decrease in demand for the bikes.