Loan data EDA
Muhammed Ahmed Abd Al ALeam
2023-09-10
Notebook Description & Setup
This is an analysis for the loan data-set on Kaggle this EDA notebook is done on the data after the data cleaning see more in the cleaning notebook here we will answer those questions below and also any question that comes in my mind or if got inspired by any viz in this notebook.
Things we will visualize:
Categorical variables distribution so I can check for data bias
Correlation pair-plot between all contentious variables
Questions:
What’s the ratio between the people with good credit score who have got their loans accepted & who got not?
How much does the credit history wellness affect the probability of getting your loan accepted?
Is there any bias of getting loan to any property area or gender in our mock company
How often does the company accept the loans per total monthly income groups.
How does the loan term correlate with the loan amount and how does this affect the loan status?
# install.packages("hrbrthemes")
# install.packages("psych")
# install.packages("waffle")
# install.packages("ggpubr")
# install.packages("DT")
# install.packages("GGally")
library(extrafont)## Registering fonts with Rlibrary(hrbrthemes)## Warning: package 'hrbrthemes' was built under R version 4.2.3## NOTE: Either Arial Narrow or Roboto Condensed fonts are required to use these themes.## Please use hrbrthemes::import_roboto_condensed() to install Roboto Condensed and## if Arial Narrow is not on your system, please see https://bit.ly/arialnarrowlibrary(scales)## Warning: package 'scales' was built under R version 4.2.3library(reshape2)## Warning: package 'reshape2' was built under R version 4.2.3library(ggpubr)## Warning: package 'ggpubr' was built under R version 4.2.3## Loading required package: ggplot2## Warning: package 'ggplot2' was built under R version 4.2.3library(tidyverse)## Warning: package 'tidyverse' was built under R version 4.2.3## Warning: package 'tibble' was built under R version 4.2.3## Warning: package 'tidyr' was built under R version 4.2.3## Warning: package 'readr' was built under R version 4.2.3## Warning: package 'purrr' was built under R version 4.2.3## Warning: package 'dplyr' was built under R version 4.2.3## Warning: package 'stringr' was built under R version 4.2.3## Warning: package 'forcats' was built under R version 4.2.3## Warning: package 'lubridate' was built under R version 4.2.3## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr 1.1.2 ✔ readr 2.1.4
## ✔ forcats 1.0.0 ✔ stringr 1.5.0
## ✔ lubridate 1.9.2 ✔ tibble 3.2.1
## ✔ purrr 1.0.1 ✔ tidyr 1.3.0## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ readr::col_factor() masks scales::col_factor()
## ✖ purrr::discard() masks scales::discard()
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errorslibrary(showtext)## Warning: package 'showtext' was built under R version 4.2.3## Loading required package: sysfonts## Warning: package 'sysfonts' was built under R version 4.2.3## Loading required package: showtextdb## Warning: package 'showtextdb' was built under R version 4.2.3##
## Attaching package: 'showtextdb'
##
## The following object is masked from 'package:extrafont':
##
## font_installlibrary(Hmisc)## Warning: package 'Hmisc' was built under R version 4.2.3##
## Attaching package: 'Hmisc'
##
## The following objects are masked from 'package:dplyr':
##
## src, summarize
##
## The following objects are masked from 'package:base':
##
## format.pval, unitslibrary(psych)## Warning: package 'psych' was built under R version 4.2.3##
## Attaching package: 'psych'
##
## The following object is masked from 'package:Hmisc':
##
## describe
##
## The following objects are masked from 'package:ggplot2':
##
## %+%, alpha
##
## The following objects are masked from 'package:scales':
##
## alpha, rescalelibrary(waffle)## Warning: package 'waffle' was built under R version 4.2.3library(GGally)## Warning: package 'GGally' was built under R version 4.2.3## Registered S3 method overwritten by 'GGally':
## method from
## +.gg ggplot2library(DT)## Warning: package 'DT' was built under R version 4.2.3COLORS <- c('#102542', '#F87060', '#CDD7D6', '#B3A394', '#FFFFFF')
FONT <- 20
PAD <- 40df <- read.csv("../data-cleaning/cleaned-data/processsed-data.csv")
datatable(data= df)summary(df)## Gender Married Education Property_Area
## Length:543 Length:543 Length:543 Length:543
## Class :character Class :character Class :character Class :character
## Mode :character Mode :character Mode :character Mode :character
##
##
##
## Monthly_Income Extra_Monthly_Income Loan_Term Credit_History
## Min. : 150 Min. : 0 Min. : 36.0 Length:543
## 1st Qu.: 2898 1st Qu.: 0 1st Qu.:104.0 Class :character
## Median : 3814 Median : 1126 Median :104.0 Mode :character
## Mean : 5352 Mean : 1547 Mean :111.8
## 3rd Qu.: 5790 3rd Qu.: 2252 3rd Qu.:104.0
## Max. :63337 Max. :33837 Max. :240.0
## Loan_Status Dependents Employment_Type Loan_Amount
## Length:543 Min. :0.0000 Length:543 Min. : 464
## Class :character 1st Qu.:0.0000 Class :character 1st Qu.:18928
## Mode :character Median :0.0000 Mode :character Median :36928
## Mean :0.7514 Mean :34896
## 3rd Qu.:2.0000 3rd Qu.:50464
## Max. :3.0000 Max. :65464
## Total_Monthly_Income
## Min. : 1442
## 1st Qu.: 4166
## Median : 5332
## Mean : 6899
## 3rd Qu.: 7546
## Max. :63337Answering the Questions
Answer 1
Let’s Start with the first question which asks about how does the credit history wellness affect the probability of getting a loan to answer this question will help us know the importance of the credit history (of course it’s important but we want to know how is it in our mock company) we will use Chi-squared & Cramer’s V tests to do this an here’s the answer:
contingency_table <- select(df, Loan_Status,Credit_History) %>%
table() %>%
as.data.frame() %>%
pivot_wider(names_from = Loan_Status, values_from = Freq) %>%
column_to_rownames(var= "Credit_History") %>%
as.matrix()
p_value <- chisq.test(contingency_table)$p.value
chi_squared <- chisq.test(contingency_table)$statistic
chi_squared_matrix <- matrix(c(chi_squared, 0, 0, chi_squared),
nrow = 2, ncol = 2, byrow = TRUE )
degrees_of_freedom <- ncol(contingency_table - 1) * nrow(contingency_table - 1)
chi_critical <- qchisq(0.05, degrees_of_freedom)
n <- sum(contingency_table)
k <- ncol(contingency_table)
r <- nrow(contingency_table)
cramers_v <- sqrt(chi_squared / (n * min(k - 1, r - 1)))
sprintf("The Credit History and the Loan Status have:\n
Chi critical = %.3f (Chi square should be higher to reject H0)
P-Value= %e (Very low)
Chi Square = %.3f
Cramer's V = %.3f\n
Which means moderate levels of relationship and we can use those numbers as a standard to compare between them to other months's because those values alone are useless.",
chi_critical, p_value, chi_squared, cramers_v) %>%
cat()## The Credit History and the Loan Status have:
##
## Chi critical = 0.711 (Chi square should be higher to reject H0)
## P-Value= 9.337448e-39 (Very low)
## Chi Square = 169.537
## Cramer's V = 0.559
##
## Which means moderate levels of relationship and we can use those numbers as a standard to compare between them to other months's because those values alone are useless.Answer 2
And here’s the Second question which asks
What’s the ratio between the people with good credit score who have got their loans accepted & who got not?
which will show us if there are any hidden problems we should investigate because the higher this number the more there are hidden problems we should dive through
chart_data <- df[df$Credit_History == 'Good', ]$Loan_Status
counts <- table(df[df$Credit_History == 'Good', ]$Loan_Status)
percentage <- (counts / length(chart_data)) * 101
chart_data <- data.frame(loan_state = names(counts),
percentage = percentage) %>% select(-c(percentage.Var1))
chart_data <- chart_data %>%
pivot_wider(names_from= loan_state, values_from= percentage.Freq)
waffle_chart <- waffle(chart_data, rows= 10,
colors= c(COLORS[2], COLORS[3]),
legend_pos= "bottom")
waffle_chart +
labs(title = 'How many people got their Loan Accepted',
subtitle= "with Credit history > 70\n") +
theme(plot.title = element_text(hjust = .5),
plot.subtitle = element_text(hjust = .5))
# ggsave("plots/getting_loan_chance_with_good_credit_history.png", width = 7, height = 7, dpi = 300)We can find that about 4/5 of the people who had good credit history got their loan accepted which also means that 20% of the applications with good credit history is rejected so we should investigate that .
Now before we go with the third question we should investigate why there are 20% of applications with good credit history are rejected so I am going to see how does the loan amount and the monthly income differs when the loans are rejected or accepted.
chart_data <- df[df$Credit_History == "Good", ]
line_y <- min(chart_data[chart_data$Loan_Status == "Accepted", ]$Total_Monthly_Income)
ggplot(chart_data, aes(x= Loan_Status, y =Total_Monthly_Income)) +
geom_jitter(colour= COLORS[3], width= 0.3) +
stat_summary(fun= median, geom= "point", shape= 18, size= 5, color= COLORS[2]) +
geom_hline(yintercept= line_y, linetype= "dashed", color= COLORS[1], size=1) +
ggtitle("Total Monlthly Income per Accepted & Rejected Loans") +
labs(y = "Total Monthly Income", x= "", subtitle= "With Credit score > 700") +
scale_y_continuous(n.breaks= 10, labels= function(y) paste0(y/1000, "K")) +
theme_classic() +
theme(plot.title = element_text(hjust = 0.5),
plot.subtitle = element_text(hjust = 0.5))## Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
## ℹ Please use `linewidth` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
# ggsave("plots/loan_acceptance_per_monthly_income.png", width = 7, height = 7, dpi = 300)Oh no we can’t find any trend at the data above except when the total monthly income was really low so now we will see if the requested Loan amount has a weird trend in the rejected loans data and if not we will investigate more.
ggplot(chart_data, aes(x= Loan_Status, y =Loan_Amount)) +
geom_jitter(colour= COLORS[3], width= 0.3) +
stat_summary(fun.y=median, geom="point", shape=15, size=5, color= COLORS[1]) +
ggtitle("Loan Amount per Accepted & Rejected Loans") +
labs(y = "Loan Amount", x= "", subtitle= "With Credit score > 700") +
scale_y_continuous(n.breaks= 10, labels= function(y) paste0(y/1000, "K")) +
theme_classic() +
theme(plot.title = element_text(hjust = 0.5),
plot.subtitle = element_text(hjust = 0.5))## Warning: The `fun.y` argument of `stat_summary()` is deprecated as of ggplot2 3.3.0.
## ℹ Please use the `fun` argument instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
# ggsave("plots/loan_amount_per_accepted_&_rejected_loans.png", width = 7, height = 7, dpi = 300)Answer 3
Now with Third question that asks about is there any bias to any category of applications of getting their loan accepted so we can identify any thing that should be investigated more in the data.
# 1st Chart Data
chart_data <- select(df, Loan_Status, Employment_Type)
chart_data$Employment_Type <- gsub("-", " ", chart_data$Employment_Type)
chart_data <- table(chart_data) %>%
as.data.frame()
# 2nd Chart Data
chart_data_filtered <- select(df[df$Credit_History == "Good", ],
Loan_Status, Employment_Type)
chart_data_filtered$Employment_Type <- gsub("-", " ", chart_data_filtered$Employment_Type)
chart_data_filtered <- table(chart_data_filtered) %>%
as.data.frame()
g1 <- ggplot(chart_data, aes(x= Employment_Type, y= Freq, fill= Loan_Status)) +
geom_bar(stat= "identity", position= "dodge") +
labs(title= "Loan Acceptableness by Employment Type",
x= "Employment type",
y= "Frequency") +
scale_fill_manual(values= c("Accepted"= COLORS[4], "Rejected"= COLORS[3])) +
theme_classic() +
theme(plot.title = element_text(hjust = 0.7),
plot.margin = margin(t= 20, b= 20, l= 20, r=20)) +
guides(fill=guide_legend(title="Loan Status"))
g2 <- ggplot(chart_data_filtered,
aes(x= Employment_Type, y= Freq, fill= Loan_Status)) +
geom_bar(stat= "identity", position= "dodge") +
labs(title= "Loan Acceptableness by Employment Type",
subtitle= "Credit History > 700",
x= "Employment type",
y= "Frequency") +
scale_fill_manual(values= c("Accepted"= COLORS[1], "Rejected"= COLORS[3])) +
theme_classic() +
theme(plot.title= element_text(hjust = 0.7),
plot.subtitle= element_text(hjust = 0.6),
plot.margin = margin(t= 20, b= 20, l= 20, r=20)) +
guides(fill=guide_legend(title="Loan Status"))
figure <- ggarrange(g1, g2,
labels = c("Non filtered", "Filtered"),
ncol = 1, nrow = 2)
figure
# ggsave("plots/loan_acceptance_per_employment_type.png", width = 7, height = 7, dpi = 300)The chart can show us important thing which is that the loan acceptableness ratio is higher in the non-self employed people when we didn’t filter the good credit histories only, but when we did the acceptableness ratio was kinda the same so we can find that employment type and in most cases all of the other categorical features affect the credit history not the loan status directly.
anyways the relationship is really low (P < 0.05) so the relationship may just be a random trend.
non_filtered_data <- select(df, Loan_Status, Employment_Type)
filtered_data <- select(df, Loan_Status, Employment_Type)[
df$Credit_History == "Good", ]
chi_squared <- chisq.test(non_filtered_data$Loan_Status,
non_filtered_data$Employment_Type)$statistic
chi_squared_filtered <- chisq.test(filtered_data$Loan_Status,
filtered_data$Employment_Type)$statistic
n <- sum(table(non_filtered_data))
k <- ncol(table(non_filtered_data))
r <- nrow(table(non_filtered_data))
n_filtered <- sum(table(filtered_data))
k_filtered <- ncol(table(filtered_data))
r_filtered <- nrow(table(filtered_data))
cramers_v <- sqrt(chi_squared / (n * min(k - 1, r - 1)))
cramers_v_filtered <-sqrt(chi_squared_filtered /
(n_filtered * min(k_filtered - 1, r_filtered - 1)))
sprintf("The correlation between Employment type and Loan status in general is:
Cramer's V= %.3f
But when we filter the data for the Good credit histories only we get:
Cramer's V= %.3f",
cramers_v, cramers_v_filtered) %>%
cat()## The correlation between Employment type and Loan status in general is:
## Cramer's V= 0.013
##
## But when we filter the data for the Good credit histories only we get:
## Cramer's V= 0.006And now with the final thing in this chart I am just going to check if there’s any effect from the living environment or the gender on the Loan status using also Cramer’s V
gender_data <- select(df[df$Credit_History == "Good", ], Loan_Status, Gender)
area_data <- select(df[df$Credit_History == "Good", ], Loan_Status, Property_Area)
married_data <- select(df[df$Credit_History == "Good", ], Loan_Status, Married)
gender_chi_squared <- chisq.test(gender_data$Loan_Status,
gender_data$Gender)$statistic## Warning in chisq.test(gender_data$Loan_Status, gender_data$Gender): Chi-squared
## approximation may be incorrectarea_chi_squared <- chisq.test(area_data$Loan_Status,
area_data$Property_Area)$statistic
married_chi_squared <- chisq.test(married_data$Loan_Status,
married_data$Married)$statistic
n_gender <- table(gender_data) %>% sum()
k_gender <- table(gender_data) %>% ncol()
r_gender <- table(gender_data) %>% nrow()
gender_cramers_v <- sqrt(gender_chi_squared /
(n_gender * min(k_gender - 1, r_gender - 1)))
n_area <- table(area_data) %>% sum()
k_area <- table(area_data) %>% ncol()
r_area <- table(area_data) %>% nrow()
area_cramers_v <- sqrt(area_chi_squared /
(n_area * min(k_area - 1, r_area - 1)))
n_married <- sum(table(married_data))
k_married <- ncol(table(married_data))
r_married <- nrow(table(married_data))
married_cramers_v <- sqrt(married_chi_squared /
(n_married * min(k_married - 1, r_married - 1)))
sprintf("The Relationship between the Gender and Loan status is:
Cramer's V= %.3f
The Relationship between the Property area and Loan status is:
Cramer's V= %.3f
The Relationship between the Marriage and Loan status is:
Cramer's V= %.3f
",
gender_cramers_v, area_cramers_v, married_cramers_v) %>%
cat()## The Relationship between the Gender and Loan status is:
## Cramer's V= 0.054
##
## The Relationship between the Property area and Loan status is:
## Cramer's V= 0.172
##
## The Relationship between the Marriage and Loan status is:
## Cramer's V= 0.113Now we can finally proof that there’s a relationship between most of
the loan requester data and the
loan status but it’s not always a direct relationship so at conclusion
we now know that:
There’s a relationship between (Gender, Property area, Marriage) and the acceptableness of the loan but it’s not always direct sometimes it may affect other variables that lead to the increase of loan acceptableness .
The self employment affects the credit history negatively sometimes that leads to decrease in the probability BUT this is very rare that makes it just a random case because the hypothesis is rejected.
Answer 4
Finally we finished the long analysis above and now let’s investigate smaller question (the fourth one) which asks: ‘How often does the company accept the loans per total monthly income groups?’.
This answer will make us see the relationship between the acceptableness ratio and the total monthly income then we can use this chart to inform our mock users with the safest minimum monthly income to apply.
and if we found that they are don’t look like a log function we will need to investigate why.
bin_labels <- c("0 ~ 2.5K", "2.5K ~ 5K", "5K ~ 7.5K", "7.5K ~ 10K", "10K ~ 12.5K",
"12.5K ~ 15K", "15K ~ 17.5K", "17.5K ~ 20K", "20K ~ 22.5K",
"22.5K ~ 25K", "25K ~ ...")
chart_data <- df %>%
mutate(Monthly_Income_Bins = cut(Total_Monthly_Income, labels = bin_labels,
breaks = c(0, 2500, 5000, 7500, 10000, 12500,
15000, 17500, 20000, 22500, 25000, Inf)))
chart_data$Loan_Status_New <- ifelse(chart_data$Loan_Status == "Accepted", 100,
ifelse(chart_data$Loan_Status == 'Rejected', 0,
chart_data$Loan_Status)) %>%
as.numeric()
chart_data <- aggregate(Loan_Status_New ~ Monthly_Income_Bins,
data= chart_data, FUN= mean)
ggplot(chart_data, aes(x= Monthly_Income_Bins, y= Loan_Status_New)) +
geom_segment(aes(x= Monthly_Income_Bins, xend= Monthly_Income_Bins,
y=0, yend= Loan_Status_New), color= COLORS[3],
size= 1.5) +
geom_point(color= COLORS[2], size=7) +
theme_classic() +
labs(title= "Loan Acceptance per Monthly Income\n",
x= "Monthly Income",
y= "Loan Acceptness") +
scale_y_continuous(n.breaks= 10, labels= function(y) paste0(y, "%")) +
theme(plot.title = element_text(hjust = 0.5),
axis.text.x = element_text(angle=90, hjust=1))
ggsave("plots/loan_acceptance_per_monthly_income.png", width = 10, height = 5, dpi = 300)As expected there’s kinda a log curve but it’s broken in two pints
which are the 25K ~ Inf and the 17.5K ~ 20K
mostly because that the records in those high ranges are rare especially
that the data is only 600 rows anyways we can find that the
sweat spot of the monthly Income is above the 2.5K also
when the monthly Income gets higher the more the requested loan amount
gets higher so as you get
higher monthly Income the more your loan acceptance depend on the other variables more.
Answer 5
Finally with the last question before starting the statistical analysis which asks :
How does the loan term correlate with the loan amount and how does this affect the loan status?
This question is crucial because it will help us know how does the loan term and the loan amount affect the loan status and if the chart was hard to understand I will use pure statistics.
ggplot(df, aes(x= Loan_Amount, y= Loan_Term, color= Loan_Status)) +
geom_line(color= COLORS[3]) +
geom_point(shape= 18, size= 3) +
theme_classic() +
scale_color_manual(values= c("Accepted"= COLORS[1], "Rejected"= COLORS[3])) +
labs(color= "Loan Status", x= "Loan amount", y= "Loan term",
title= "Loan amount & term affect on the Loan Acceptance\n") +
scale_x_continuous(n.breaks= 8, labels= function(x) paste0(x/1000, "K")) +
scale_y_continuous(n.breaks= 6, labels= function(y) paste0(y, " Day")) +
theme(legend.position = "bottom", legend.background= element_rect("#f0f0f0"),
plot.title = element_text(hjust = 0.5))
# ggsave("plots/acceptance_per_loan_amount_&_term.png", width = 10, height = 7, dpi = 300)We can find really important insight with this chart only which is :
our mock company usually doesn’t accept the loans with high
terms nor low ones they usually accept loan term of
120 Days so our mock company should create special loans
for people who wants long or short terms even those people are kinda
rare and accept wider range of credit histories in terms of long
loans.
Boring statistical charts
those charts may not give us many insights as the ones before but they still important to learn us more about our data and I used them to modify some of the insights above to get more accurate insights.
anyways I didn’t write any insights here only the charts and they look ugly because R sucks at pair-plots.
chart_data <- select(df, Loan_Amount, Loan_Term,
Total_Monthly_Income, Loan_Status)
ggpairs(chart_data, columns= 1:3, ggplot2::aes(colour= Loan_Status),
upper = list(continuous= "cor"),
lower = list(continuous= "points", combo = "dot_no_facet", color = "Loan_Status")) +
scale_y_continuous(
n.breaks = 5,
labels = function(y) {
ifelse(y >= 1000, paste0(y / 1000, "K"), y)}) +
scale_x_continuous(
n.breaks = 5,
labels = function(x) {
ifelse(x >= 1000, paste0(x / 1000, "K"), x)}) +
theme_bw() +
scale_fill_manual (values = c(COLORS[2], COLORS[3])) +
scale_color_manual (values = c(COLORS[2], COLORS[3])) +
labs(title= "Loan Features Correlation Pair plot\n") +
theme(plot.title = element_text(hjust = 0.5))## Scale for y is already present.
## Adding another scale for y, which will replace the existing scale.
## Scale for y is already present.
## Adding another scale for y, which will replace the existing scale.
## Scale for y is already present.
## Adding another scale for y, which will replace the existing scale.
## Scale for y is already present.
## Adding another scale for y, which will replace the existing scale.
## Scale for y is already present.
## Adding another scale for y, which will replace the existing scale.
## Scale for y is already present.
## Adding another scale for y, which will replace the existing scale.
## Scale for x is already present.
## Adding another scale for x, which will replace the existing scale.
## Scale for x is already present.
## Adding another scale for x, which will replace the existing scale.
## Scale for x is already present.
## Adding another scale for x, which will replace the existing scale.
# ggsave("plots/loan_features_corr_paiplot.png", width = 7, height = 7, dpi = 300)getmode <- function(v) {
uniqv <- unique(v)
uniqv[which.max(tabulate(match(v, uniqv)))]
}
chart_data <- select(df, Married, Education, Property_Area, Loan_Status) %>%
tibble::rowid_to_column("Loan_ID") %>%
melt(id.vars="Loan_ID") %>%
group_by(variable) %>%
mutate(max_frequency = value == getmode(value))
ggplot(chart_data, aes(Loan_ID, value, fill= max_frequency)) +
geom_bar(stat = "identity") +
stat_smooth() +
facet_wrap(~variable, , scales = "free") +
theme_classic() +
labs(x= NULL, y= NULL, title= "Categorical columns distrbution\n") +
scale_fill_manual(values = c(COLORS[3], COLORS[4])) +
guides(fill = FALSE) +
scale_x_continuous(
n.breaks = 5,
labels = function(x) {
ifelse(x >= 1000, paste0(x / 1000, "K"), x)}) +
theme(plot.title = element_text(hjust = 0.5))## Warning: The `<scale>` argument of `guides()` cannot be `FALSE`. Use "none" instead as
## of ggplot2 3.3.4.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.## `geom_smooth()` using method = 'loess' and formula = 'y ~ x'
# ggsave("plots/categorical_cols_dist.png", height= 7, width= 8 , dpi= 300)Conclusions
Finally we finished this EDA it may not be the best visualization I’ve ever made any ways now I will share with you some conclusions but firstly read this note first:
This whole analysis is done on MOCK DATA so DON’T use the project insights nor the ML in anything related to business nor money at general and anything you’ll do with this insights is on your own risk but you MUSTN’T DO THAT!!
Now here are some conclusions from our EDA:
We can find that the Total monthly income and the loan acceptance ratio have log curve which means that the Loan acceptance ratio doesn’t get affected after 2.5K of Total monthly income.
High loan terms and low terms has slightly more more probability to get rejected.
About 20% of the loans request with good credit histories get rejected (mostly because the higher loan term or because low or high loan request)
That’s enough for the conclusions phase for more analysis go read the detailed analysis for each chart.