Telecom Case Study: Churn Prediction

Dashboard presents results of our model regressor which predicts whether a customer will churn or not, for a typical telecommunications provider.
Following ingestion from multiple sources, we explore multi-characteristics of the data, build our model and finally present the results.

Data exhibiiting multi-characteristic distributions

Internet Service Distribution

Fiber Optic

DSL

No Internet

Gender Distribution in Customer Attrition

Churn

Non Churn

Customer Attrition Data

Customer attrition tenure group

Churn

Non Churn

Monthly Charge Distribution in Customer attrition

Churn

Non Churn

Ingestion of the data from multiple data sources

Source 1: Demographics Data

customerID	gender	SeniorCitizen	Partner	Dependents
7590-VHVEG	Female	No	Yes	No
5575-GNVDE	Male	No	No	No
3668-QPYBK	Male	No	No	No
7795-CFOCW	Male	No	No	No
9237-HQITU	Female	No	No	No
9305-CDSKC	Female	No	No	No
1452-KIOVK	Male	No	No	Yes
6713-OKOMC	Female	No	No	No
7892-POOKP	Female	No	Yes	No
6388-TABGU	Male	No	No	Yes
9763-GRSKD	Male	No	Yes	Yes
7469-LKBCI	Male	No	No	No
8091-TTVAX	Male	No	Yes	No
0280-XJGEX	Male	No	No	No
5129-JLPIS	Male	No	No	No
3655-SNQYZ	Female	No	Yes	Yes
8191-XWSZG	Female	No	No	No
9959-WOFKT	Male	No	No	Yes
4190-MFLUW	Female	No	Yes	Yes
4183-MYFRB	Female	No	No	No
8779-QRDMV	Male	Yes	No	No
1680-VDCWW	Male	No	Yes	No
1066-JKSGK	Male	No	No	No
3638-WEABW	Female	No	Yes	No
6322-HRPFA	Male	No	Yes	Yes
6865-JZNKO	Female	No	No	No
6467-CHFZW	Male	No	Yes	Yes
8665-UTDHZ	Male	No	Yes	Yes
5248-YGIJN	Male	No	Yes	No
8773-HHUOZ	Female	No	No	Yes
3841-NFECX	Female	Yes	Yes	No
4929-XIHVW	Male	Yes	Yes	No
6827-IEAUQ	Female	No	Yes	Yes
7310-EGVHZ	Male	No	No	No
3413-BMNZE	Male	Yes	No	No
6234-RAAPL	Female	No	Yes	Yes
6047-YHPVI	Male	No	No	No
6572-ADKRS	Female	No	No	No
5380-WJKOV	Male	No	No	No
8168-UQWWF	Female	No	No	No
8865-TNMNX	Male	No	Yes	Yes
9489-DEDVP	Female	No	Yes	Yes
9867-JCZSP	Female	No	Yes	Yes
4671-VJLCL	Female	No	No	No
4080-IIARD	Female	No	Yes	No
3714-NTNFO	Female	No	No	No
5948-UJZLF	Male	No	No	No
7760-OYPDY	Female	No	No	No
7639-LIAYI	Male	No	No	No
2954-PIBKO	Female	No	Yes	Yes
8012-SOUDQ	Female	Yes	No	No
9420-LOJKX	Female	No	No	No
6575-SUVOI	Female	Yes	Yes	No
7495-OOKFY	Female	Yes	Yes	No
4667-QONEA	Female	Yes	Yes	Yes
1658-BYGOY	Male	Yes	No	No
8769-KKTPH	Female	No	Yes	Yes
5067-XJQFU	Male	Yes	Yes	Yes
3957-SQXML	Female	No	Yes	Yes
5954-BDFSG	Female	No	No	No
0434-CSFON	Female	No	Yes	No
1215-FIGMP	Male	No	No	No
0526-SXDJP	Male	No	Yes	No
0557-ASKVU	Female	No	Yes	Yes
5698-BQJOH	Female	No	No	No
5122-CYFXA	Female	No	No	No
8627-ZYGSZ	Male	No	Yes	No
3410-YOQBQ	Female	No	No	No
3170-NMYVV	Female	No	Yes	Yes
7410-OIEDU	Male	No	No	No
2273-QCKXA	Male	No	No	No
0731-EBJQB	Female	No	Yes	Yes
1891-QRQSA	Male	Yes	Yes	Yes
8028-PNXHQ	Male	No	Yes	Yes
5630-AHZIL	Female	No	No	Yes
2673-CXQEU	Female	Yes	No	No
6416-JNVRK	Female	No	No	No
5590-ZSKRV	Female	No	Yes	Yes
0191-ZHSKZ	Male	Yes	No	No
3887-PBQAO	Female	No	Yes	Yes
5919-TMRGD	Female	No	No	Yes
8108-UXRQN	Female	No	Yes	Yes

Source 2: Subscription Data

Source 3: Churn Data

customerID	Churn
7590-VHVEG	No
5575-GNVDE	No
3668-QPYBK	Yes
7795-CFOCW	No
9237-HQITU	Yes
9305-CDSKC	Yes
1452-KIOVK	No
6713-OKOMC	No
7892-POOKP	Yes
6388-TABGU	No
9763-GRSKD	No
7469-LKBCI	No
8091-TTVAX	No
0280-XJGEX	Yes
5129-JLPIS	No
3655-SNQYZ	No
8191-XWSZG	No
9959-WOFKT	No
4190-MFLUW	Yes
4183-MYFRB	No
8779-QRDMV	Yes
1680-VDCWW	No
1066-JKSGK	Yes
3638-WEABW	No
6322-HRPFA	No
6865-JZNKO	No
6467-CHFZW	Yes
8665-UTDHZ	Yes
5248-YGIJN	No
8773-HHUOZ	Yes
3841-NFECX	No
4929-XIHVW	No
6827-IEAUQ	No
7310-EGVHZ	No
3413-BMNZE	No
6234-RAAPL	No
6047-YHPVI	Yes
6572-ADKRS	No
5380-WJKOV	Yes
8168-UQWWF	Yes
8865-TNMNX	No
9489-DEDVP	No
9867-JCZSP	No
4671-VJLCL	No
4080-IIARD	No
3714-NTNFO	No

Algorithm & Model Building

Feature Importance

Churn

Non Churn

Confusion Matrix

Churn

Non Churn

Churn

Non Churn

Churn

278

212

1158

110

Receiver Operating Characteristic

AUC - 0.79

Model Results

Test Table

customerID	Actual Churn Prob.	Non-Churn Prob.	Churn Prob.	Predicted Churn Prob.
0871-URUWO	Yes	0.323	0.677	Yes
9497-QCMMS	Yes	0.006	0.994	Yes
4229-CZMLL	Yes	0.356	0.644	Yes
0488-GSLFR	Yes	0.031616667	0.968383333	Yes
6358-LYNGM	Yes	0.23	0.77	Yes
1062-LHZOD	Yes	0.35	0.65	Yes
6651-AZVTJ	Yes	0.352	0.648	Yes
5275-SQEIZ	Yes	0.137	0.863	Yes
3727-RJMEO	Yes	0.225	0.775	Yes
5624-BQSSA	Yes	0.372034524	0.627965476	Yes
9369-XFEHK	Yes	0.301	0.699	Yes
7228-OMTPN	Yes	0.363	0.637	Yes
4800-VHZKI	Yes	0.247692063	0.752307937	Yes
9185-TQCVP	Yes	0.246	0.754	Yes
4706-AXVKM	Yes	0.273	0.727	Yes
7130-VTEWQ	Yes	0.382	0.618	Yes
8185-UPYBR	Yes	0.232	0.768	Yes
7511-YMXVQ	Yes	0.364	0.636	Yes
2344-JMOGN	Yes	0.287	0.713	Yes
4910-GMJOT	Yes	0.045	0.955	Yes
9878-TNQGW	Yes	0.363	0.637	Yes
6661-HBGWL	Yes	0.447	0.553	Yes
6688-UZPWD	Yes	0.141	0.859	Yes
7593-XFKDI	Yes	0.488	0.512	Yes
0612-RTZZA	Yes	0.184	0.816	Yes
6614-YWYSC	No	0.998	0.002	No
9546-KDTRB	No	0.798	0.202	No
5151-HQRDG	No	0.946	0.054	No
6624-JDRDS	No	0.779	0.221	No
3082-VQXNH	No	0.557	0.443	No
1309-XGFSN	No	0.837	0.163	No
8663-UPDGF	No	0.853	0.147	No
0455-ENTCR	No	0.962	0.038	No
4361-FEBGN	No	1	0	No
3898-BSJYF	No	0.774	0.226	No
8869-LIHMK	No	0.922	0.078	No
5482-VXSXJ	No	0.59	0.41	No
8587-XYZSF	No	0.967	0.033	No
3551-GAEGL	No	0.924	0.076	No
7463-IFMQU	No	1	0	No
5404-GGUKR	No	0.671	0.329	No
7394-LWLYN	No	0.618983333	0.381016667	No
2027-DNKIV	No	0.97	0.03	No
4213-HKBJO	No	0.994	0.006	No
1915-OAKWD	No	0.948	0.052	No
3946-JEWRQ	No	0.763	0.237	No
8041-TMEID	No	0.903	0.097	No
5863-OOKCL	No	0.719	0.281	No
2740-JFBOK	No	0.899	0.101	No
2823-LKABH	No	0.746	0.254	No
9402-ORRAH	No	0.303	0.697	Yes
2038-OEQZH	No	0.458	0.542	Yes
1751-NCDLI	No	0.061	0.939	Yes
0432-CAJZV	No	0.384	0.616	Yes
5348-CAGXB	No	0.391	0.609	Yes
7401-RUBNK	No	0.471	0.529	Yes
2103-ZRXFN	No	0.435	0.565	Yes
3088-LHEFH	No	0.496	0.504	Yes
9290-SHCMB	No	0.212	0.788	Yes
4706-DGAHW	No	0.267	0.733	Yes
9500-LTVBP	No	0.444	0.556	Yes
4367-NHWMM	No	0.259	0.741	Yes
2135-DQWAQ	No	0.345	0.655	Yes
7952-OBOYL	No	0.135	0.865	Yes
1077-HUUJM	No	0.379	0.621	Yes
8040-MNRTF	No	0.002	0.998	Yes
2465-BLLEU	No	0.377616667	0.622383333	Yes
4537-DKTAL	No	0.446	0.554	Yes
0410-IPFTY	No	0.416	0.584	Yes
7901-HXJVA	No	0.464	0.536	Yes
3453-RTHJQ	No	0.435	0.565	Yes
1932-UEDCX	No	0.389	0.611	Yes
5119-KEPFY	No	0.449	0.551	Yes
7126-RBHSD	No	0.425	0.575	Yes
8990-ZXLSU	No	0.383	0.617	Yes
7993-PYKOF	Yes	0.608	0.392	No
1902-XBTFB	Yes	0.62	0.38	No
9614-RMGHA	Yes	0.765	0.235	No
4640-UHDOS	Yes	0.682	0.318	No
4106-HADHQ	Yes	0.918	0.082	No
9048-JVYVF	Yes	0.704	0.296	No
8818-XYFCQ	Yes	0.67	0.33	No
6311-UEUME	Yes	0.604	0.396	No
8580-QVLOC	Yes	0.932	0.068	No
2070-FNEXE	Yes	0.595	0.405	No
3018-TFTSU	Yes	0.623	0.377	No
6352-TWCAU	Yes	0.555	0.445	No
4174-LPGTI	Yes	0.939	0.061	No
3045-XETSH	Yes	0.536	0.464	No
0324-BRPCJ	Yes	0.837	0.163	No
8436-BJUMM	Yes	0.652	0.348	No
0117-LFRMW	Yes	0.841	0.159	No
2817-NTQDO	Yes	0.752	0.248	No
3058-HJCUY	Yes	0.689	0.311	No
5692-FPTAH	Yes	0.715	0.285	No
8740-XLHDR	Yes	0.662	0.338	No
0637-UBJRP	Yes	0.622	0.378	No
0125-LZQXK	Yes	0.51	0.49	No
9578-VRMNM	Yes	0.736	0.264	No

Probability Prediction on Top 10 Test Samples

Churn

Non Churn

Management Recommendations

1. Retaining high profitable customers is the number one business goal.
2. Companies need to predict which customers are at high risk of churn.
3. We recommend the company to consider Features Importance chart which are strong indicators to manage customer churn.