Audio Recording May 8, 2022 at 4:49:54 PM. dATA aNALYSIS OF aVOCADOS. 

Scene 1

Audio Recording May 8, 2022 at 4:49:54 PM

Audio Recording May 8, 2022 at 4:52:56 PM. One in a crowd. 

Scene 2

Lawrence Abhinav  Gunkonda Carolina  Poyekar Melina Sharma  Anuja  Rai

Audio Recording May 8, 2022 at 4:52:56 PM

Audio Recording May 8, 2022 at 4:57:32 PM. Project Motivation. 

Scene 3

Our project will tackle this avocado demand in this dataset.  Our project is mainly focusing on predicting the price of avocados based on different factors such as days, seasons and so on.

Audio Recording May 8, 2022 at 4:57:32 PM

Audio Recording May 8, 2022 at 5:00:09 PM. Project Aim. 

Scene 4

Can prediction of the price of avocados be done based on date, total volume, bags year, type, and region?  We are also predicting  the sales volume of avocados

Audio Recording May 8, 2022 at 5:00:09 PM

Audio Recording May 8, 2022 at 5:01:52 PM. Audio Recording May 8, 2022 at 5:03:54 PM. 

Scene 5

The dataset consists of 18249 rows × 13 columns. As far as the data is concerned, we noticed there were no null values  The data was reported on the weekly sale of avocado from 2015 to the quarter of 2018  Type and Region were of categorical data

E]:
myDF . isnull() . sum()
Date
AveragePrice
Total Volume
4 e46
4225
4770
Total Bags
Small Bags
Large Bags
XLarge Sags
type
year
region
dtype :
intS4

Audio Recording May 8, 2022 at 5:01:52 PM

Audio Recording May 8, 2022 at 5:03:54 PM

Audio Recording May 8, 2022 at 5:05:15 PM

Audio Recording May 8, 2022 at 5:06:13 PM. Distribution of the data. 

Scene 6

The describe function indicates the statistical summary, whether the data is evenly distributed. To visualize this, we decided to plot it in a histogram. The Average Price is skewed to the right

Audio Recording May 8, 2022 at 5:06:13 PM

Audio Recording May 8, 2022 at 5:07:54 PM. Data Transformation. 

Scene 7

The above columns were used as our ‘x’s The columns that were excluded were PLU (4046, 4225, 4770) and Total Bags since the data was redundant.

Audio Recording May 8, 2022 at 5:07:54 PM

Audio Recording May 8, 2022 at 5:09:47 PM. Models and Analysis. 

Scene 8

ADF statistic: -2.096522
"-value: 0.245910
Critical Values:
-3.472
5%: -2.880
10%: -2.577
series is non-stationary
Root Nean squared Error: 0.10541802878335570

Stationary vs Non – Stationary Arima Model In our studies, we encountered with the unclear difference between stationary and non-stationary attribute of our dataset due to which we had drop Arima model and choose Linear Regression Model.

Audio Recording May 8, 2022 at 5:09:47 PM

Audio Recording May 8, 2022 at 5:12:51 PM. Linear Regression Model Support Vector Machine. 

Scene 9

Linear Regression Model  Support Vector Machine

We filtered with the variables that we wanted to use. Our target was the “Average Price” and for our “X” we decide to use the rest of the columns and dropped our “Average Price” column.

Audio Recording May 8, 2022 at 5:12:51 PM

Audio Recording May 8, 2022 at 5:14:56 PM. Train and tested the data with a Linear Regression and a SVM model.. 

Scene 10

Train and tested the data with a Linear Regression and a SVM model.

X train, X_test, y _ train, y _ test = y,
["Linear Regression" , "Support Vector Machine"
alg_names =
) svm. SVR( ) ]
alg_models =
for in alg_models):
print( "For algorithm" , alg_name)
alg_mode1.fit(X_train,y_train)
predictions =
alg_model . predict (X _ test)
print( "Mean absolute error for" predictions))
print("Mean squared error for" squared_error(y_test, predictions))

Audio Recording May 8, 2022 at 5:14:56 PM

Audio Recording May 8, 2022 at 5:17:29 PM. The output indicated that Linear Regression Model was the best model with the lowest score of Mean Absolute Error as 0.205 and Mean Squared Error of 0.75 as compared to the Support Vector Machine which has MAE of 0.257 and RMSE of 0.113. The test size is 0.25 .. 

Scene 11

The output indicated that Linear Regression Model was the best model with the lowest score of Mean Absolute Error as 0.205 and Mean Squared Error of 0.75 as compared to the Support Vector Machine which has MAE of 0.257 and RMSE of 0.113. The test size is 0.25 .

Audio Recording May 8, 2022 at 5:17:29 PM

Audio Recording May 8, 2022 at 5:18:03 PM. Audio Recording May 8, 2022 at 5:23:30 PM. 

Scene 12

Plotting the actual and predictions: this drew further analysis to reduce the number of attributes in the dataset.

Compare Actual and Predictions
30 -
20 -
15 -
00
05
10
20
ktual
30

Audio Recording May 8, 2022 at 5:18:03 PM

Audio Recording May 8, 2022 at 5:23:30 PM

Audio Recording May 8, 2022 at 5:23:58 PM. Turned the Types and Region into numerical data.. 

Scene 13

Turned the Types and Region into numerical data.

Audio Recording May 8, 2022 at 5:23:58 PM

Audio Recording May 8, 2022 at 5:19:55 PM. Linear Regression vs Support Vector. 

Scene 14

Compare Actual and Predictions
30-
25 -
20 -
15 -
10 -
00
10
15
ktual
20
30

In this case the RMSE SVM did a bit better. Since SVM is a bit more complicated we decided to stick with the Linear Regression.

Audio Recording May 8, 2022 at 5:19:55 PM

Audio Recording May 8, 2022 at 5:20:52 PM. Findings. 

Scene 15

The slope represents the prediction and the dots around the line represents the actual value of the price of the Avocados. The fact that the actual values are around the prediction and the lower error tells that Linear Regression model should be chosen to predict the price of Avocado.

Compare Actual and Predictions
30 -
20
E 15 -
05 -
00
05
20
ktual
30

Audio Recording May 8, 2022 at 5:20:52 PM

Audio Recording May 8, 2022 at 5:21:35 PM. Conclusion. 

Scene 16

This project was initiated with an objective to predict the weekly avocado price based on various models relevant to the given dataset.  In terms of finding the correlation of price with other variables, three models, namely Arima, Linear Regression and Support Vector machine were applied.  After training and testing the dataset of Avocado with these models, it is found that Linear Regression Model is the champion model to predict the prices because it gave the least amount of error in the price predictions from the actual dataset.

Audio Recording May 8, 2022 at 5:21:35 PM