5.1 Implementation Platform
1.Hardware
1. Processor:Intel Duo Core
2. RAM: 8 GB
3. GPU: NVIDIA Tesla K80
2.Software
1. Operating System:Linux
2. Programming Languages: Python
3. Server: Reddis
5.2 Postulatesset Introduction
The postulatesset used in this product is from the
rst year (Fevery 2012, Spring 2013, and
Summer 2013 semesters) of MITx and HarvardX roads o ered on the edX
platform. The postulatesset includes 641138 archives, and each annals resembles sev-
eral learner’s activities in a road. In each annals, we convergence on the information
showed in the
gure 5 beneath. The postulatesset consists of learners’ principle postulates in
11 completed roads. There are 57400 archives in each of which learner’s measure
is inoperative, so these archives are deleted.
Dept. of CSE, DSCE, Bangalore 78 Page 16
Predictions Of Dropouts In MOOCs 2015-19
Figure 5: Fickles in each annals of postulatesset.
Based on Chapters fickle, every learners can be disconnected into three throngs
including merely registered, open, and locomotive. The throng of merely registered
represents the learners who never path roadware, the throng of open repre-
sents the learners who path roadware beside path short than half of the beneficial
courseware chapters, and the throng of locomotive resembles the learners who path
aid than half of the beneficial roadware chapters.
Nearly 37%, 57% and 6%
of the learners suit to the throngs of merely registered, open and locomotive re-
spectively. So most of the learners are referable attributable attributable attributable zealous coercion education in MOOCs,
and merely scant enjoy learnt most road full.
5.3 Comportment Dissection
Both measure and certi
cate fickle are material indicator coercion evaluating learn-
ing e ect coercion learners in MOOCs. So we admit a statistic on measure to pretence the
dispensation of learners in 11 roads. From the statistic we can recognize that learn-
ers can be part-among into three categories in each road, relish most of the learners
with measure of 0, encircling 10%-20% learners with measure balance 0, beside didn’t earn
certi
cate in road, and the proportion of learners who earned certi
cate is encircling
3%-10%. Coercion retirement, we cevery the instances of measure=0, measure>0 beside no cer-
Dept. of CSE, DSCE, Bangalore 78 Page 17
Predictions Of Dropouts In MOOCs 2015-19
ti
cated, and certi
cated as three categories relish systematize 0, systematize 1 and systematize 2
respectively.
Based on the aggravate dissection, we examine to comprehend the di erence among
the three categories. We congenial the balance, stint, pity quantile, half
quantile, three pitys quantile, consummation coercion learner’s comportment elements relish
events, days, videos, chapters, coercionum respectively. Based on these statistics,
we can perceive-keep that the learners who gain elevated measure are aid locomotive than the
learners with deep measure in MOOCs. Coercion specimen, coercion the instance of systematize 0, the
balance of incident, days, videos, and chapters are respectively entire 100, 3, 17 and
1, beside coercion the instance of systematize 1 the balance gum are respectively larger than 1000,
13, 100, and 5, and coercion the instance of systematize 2 the balance gum are respectively
larger than 5000, 40, 400 and 10.
In manage to comprehend the dispensation of the learners with di erent cate-
gories in comportment element extension. We applied K-media algorithm to the learners’
comportment elements relish Incidents, Days, Videos, Chapters, Coercionum in Table 1 coercion a
course. Every the learners are thronged into couple throngs or three throngs, and
then we congenial the balance, stint, pity quantile, half quantile, three
quarters quantile, consummation coercion learners’ measure in each throng pretenceed in Table
2 and Table 3.
From the results in Table 2, we can con
dently conjecture that the throng A and
the throng B can resemble the learners who enjoy a faulty act or amend
act in online education respectively. And aid the throng A and the
throng B can be guarded as the learners with the measure of 0 and the learners
with measure balance 0 respectively. The mediocre illimitableness among the points of the
throng A and the throng B is 6.23 in comportment element extension.
From the results in Table 3, we can perceive-keep that the throng A’ and the
throng C’ can be guarded as systematize 0 or systematize 2 respectively. The throng B’
includes the archives end from the throng A and the throng B in Table 2 and
represents the learners whose education act openly and enjoy medium
measure in criterion. So we can affect the throng B’ in Table 3 resembles systematize
1. The mediocre illimitableness among the points of the throng A’ and the throng
B’, the throng B’ and the throng C’, the throng A’ and the throng C’ is 5.91,
2.72 and 7.87 respectively in comportment element extension. It balances that there are
overlapping among systematize 1 and systematize 2, and amend separability among systematize 0
